Intensive Care Med (2001) 27: 1064±1072 DOI 10.1007/s001340100951

Markus Weiss Joachim Fischer Monika Boeckmann Brigitte Rohrer Oskar Baenziger

Received: 8 December 2000 Final revision received: 8 December 2000 Accepted: 21 February 2001 Published online: 16 May 2001  Springer-Verlag 2001

)

M. Weiss ( ) ´ J. Fischer ´ M. Boeckmann ´ B. Rohrer ´ O. Baenziger Department of Intensive Care and Neonatology, University Children's Hospital, Steinwiesstrasse 75, 8032 Zurich, Switzerland E-mail: [email protected] Phone: +41-1-2 66 71 11 Fax: +41-1-2 66 71 68

NE ON ATA L A ND PE DI ATR IC IN TENS IVE CA RE

Evaluation of a simple method for minimizing iatrogenic blood loss from discard volumes in critically ill newborns and children

Abstract Objective: To validate a simple method avoiding discard volumes in pediatric patients with indwelling arterial and venous lines. Design: Zero-discarding was achieved by passive extracorporeal arteriovenous backflow via standard single pressure transducer equipment. We tested backflow distances (10, 20 and 30 cm beyond the sampling port), corresponding to withdrawal volumes of 0.6 ml, 0.8 ml and 1.0 ml, respectively, in comparison to conventional sampling with discard of 0.6 ml. With the backflow technique, the ªwithdrawal volumeº was flushed back to the patient after sampling. We enrolled 120 patients who were allocated to either of the following paired sampling procedures: 10 cm versus conventional, 20 cm versus conventional, 30 cm versus conventional and two paired conventional samples. The order of the sampling was randomly allocat-

Introduction The measurement of blood gases, electrolytes and other blood compounds is a cornerstone of critical care. Indwelling arterial or central venous vascular catheters facilitate on-demand withdrawal of blood samples. In small infants, iatrogenic blood loss due to repetitive sampling may lead to anemia and may increase the need for blood transfusions, with its associated risk of adverse reactions and infections [1, 2, 3]. When blood is collected from indwelling catheters, a certain amount of blood is usually withdrawn and dis-

ed. Bias and precision were determined using Bland-Altman diagrams and algorithms. Results: No appreciable difference was found for blood gases, hemoglobin, potassium and calcium between the backflow technique and conventional sampling. Sodium results and blood glucose showed a bias towards higher values with the backflow technique (mean difference, sodium, 0.9 mmol/l; glucose, mean difference 0.5 mmol/l, standard deviation 0.44 mmol/l). Conclusions: The backflow technique provides reliable results for blood gases and electrolytes. However, in patients at risk of hypoglycemia, the backflow method should not be used to monitor blood glucose levels. Keywords Arterial blood sampling ´ Blood conserving ´ Invasive pressure monitoring ´ Pediatric, neonatal intensive care

carded before obtaining the sample, in order to minimize contamination with the catheter flushing solution. It has been estimated that the volume of this discarded blood accounts for 24±30 % of the total amount of blood withdrawn for diagnostic sampling [1, 4]. Some clinicians prefer to return the withdrawn deadspace volume to the patient. This practice, however, bears the potential risk of arterial embolism (air, clots) and contamination with infectious agents. During the past decade, several measures aimed at minimizing iatrogenic blood loss have been introduced. These include the use of pediatric-sized laboratory col-

1065

Fig. 1 Schema of the bloodconserving arterial line system configuration. If the two stopcocks (C) of the device are set appropriately as indicated in the figure, the device allows passive backflow of fluid (arrows) from the arterial pressure line (F) into the central venous pressure line (K), driven by the arteriovenous pressure difference. Further letters indicate: A = double stopcock device; B = blood pressure transducer, C = intravenous infusion, D = flush system, E = stopcocks, F = arterial pressure line, G = sampling stopcock, H = pediatric pressure tubing, I = arterial cannula, K = central venous pressure line, L = central venous catheter

lections tubes, batching the requests for laboratory tests and monitoring the cumulative volume of blood taken from individual patients [1, 4, 5, 6, 7]. To reduce the need for erythrocyte transfusions, the application of recombinant human erythropoietin has been suggested [8, 9]. More recent strategies to minimize blood loss include the use of sensors and electromechanical devices to evaluate blood chemistries and arterial blood gases ªin lineº [10, 11, 12, 13, 14, 15, 16], and the use of blood-conserving arterial blood lines [17, 18, 19, 20, 21, 22, 23]. From the latter, only the blood required for diagnostic testing is withdrawn, while the volume usually discarded is returned to the patient. These blood-conserving arterial blood sampling systems include a capacitive element and a volume-restricted reservoir for aspiration of blood from the patient. The use of these systems in neonatal and pediatric intensive care is hampered by the high volume which is drawn back into the system and the negative pressures generated by the system during aspiration. In our own unit the latter frequently led to temporary collapse of the artery vessel. Moreover, flushing the capacitive element free from remaining blood required large quantities of flushing solution, a particular disadvantage in infants with fluid restriction. These problems forced us to search for a simple method for the closed return of deadspace volumes. The principle of withdrawing the usual discard volume via the arterial line for subsequent reinfusion can also be realized by means of a single transducer system for continuous arterial and intermittent central venous pressure measurement, which were originally proposed, in 1979, by Rhoads and Kariman and by Calisi and Wei-

gelt in 1983 [24, 25]. In our unit, such a double-stopcock system is employed for continuous arterial blood pressure monitoring with intermittent readings of the central venous pressure. With this system extracorporeal passive arteriovenous blood flow from the arterial cannula to the central venous catheter can also be established (Fig. 1). This feature can be used to withdraw the required deadspace volume into the proximal arterial line by passive backflow. Subsequently, blood may be withdrawn from the arterial sampling port or stopcock. After completion of blood sampling, the discard volume is flushed back to the patient through the arterial cannula. The proposed method does not incur additional cost or waste. The purpose of the present study was to compare systematically the results of the arterial backflow technique with the results obtained from the conventional, discarding procedure.

Material and methods Patients and setting We enrolled 120 critically ill newborns and children who were admitted to a 19-bed tertiary multidisciplinary intensive care unit. Patients were included when an arterial and a central venous catheter were present, and free flow of blood could be obtained from the arterial cannula for blood sampling. Premature infants were excluded, as well as patients with hemodynamic instability or patients receiving vasoactive drugs through the venous line used for the backflow technique. The study was approved by the institutional review board. The board waived the need for informed consent.

1066

Study design We compared the results from samples obtained by the conventional technique with samples obtained by the backflow technique using three different backflow distances. All paired samples were withdrawn from the patient within 4 min. The order of sampling (conventional vs backflow technique) was randomly allocated. Each series included 30 patients, only one set of samples was obtained from each patient. The resulting comparisons were: conventional versus a backflow distance of 10 cm; conventional versus a backflow distance of 20 cm and conventional versus a backflow distance of 30 cm. The series were completed in the following order: 30 cm, 20 cm, 10 cm. For comparing the bias and precision of the backflow technique with the conventional method, we considered the following control measurements as appropriate ªgold standardº: first, two consecutive samples were obtained by the conventional method. These controls are referred to as ªconventional versus conventionalº. Second, we performed duplicate measurements of randomly selected 90 samples from all series. These duplicate measurements elucidate the precision and bias attributable to storage and re-test reliability of the analyzer when re-analyzing the same sample. All samples were entered into the analyzer (ABL, Radiometer, Kopenhagen, Denmark) within 1 min of collection. The following eight parameters were determined: blood gases (pH, PO2, CO2), electrolytes (Na, K, ionized Ca), hemoglobin and blood glucose. Study procedures Figure 1 illustrates the setup for all measurements: The doublestopcock pressure transducer equipment (Cath Lab Set, Homedica, Cham, Switzerland) includes an arterial and a venous pressure line, each 1.75 m in length. The arterial pressure line was connected to the sampling stopcock distally followed by a 10-cm infusion line (PE-infusion line, Clinico, Homburg, Germany), which was connected directly to the indwelling arterial cannula. The distal end of the venous pressure line was connected to two stopcocks and a 10-cm infusion line (PE-infusion line, Clinico, Homburg, Germany), which provided the connection to the central venous catheter. The pressure transducer and the arterial pressure line were flushed with heparinized (1 IU heparin/ml) Ringer's solution (Na 131 mmol/l, K 5.4 mmol/l, Ca 1.8 mmol/l; Hartmann Ringersolution, Bioren, Couvet, Switzerland). Diagnostic arterial blood samples were drawn using the conventional or the passive arteriovenous backflow technique as follows: · For the conventional technique, the distal sampling stopcock was opened, 0.6 ml of discard volume was withdrawn into a standard 2-ml syringe, followed by withdrawal of a further 0.6 ml of blood for diagnostic testing into a preheparinized syringe (PICO 30, Radiometer, Copenhagen, Denmark). Afterwards, the remaining blood was flushed from the stopcock outlet first, and then the line was flushed. · With the backflow technique, blood was collected as follows: Marks were attached at 10, 20 or 30 cm proximal to the sampling stopcock. With the pressure line used (the same type in all patients), a length of 10 cm corresponds to an internal volume of 0.2 ml and to a total withdrawn volume of 0.6 ml (volume of arterial cannula, stopcock and 10 cm pressure tubing included). Before connecting the sampling syringe to the sampling port, we evacuated any remaining Ringer's solution from the sampling stopcock outlet. Then the preheparinized syringe was attached to the port and the double-stopcocks on the pressure

transducer were opened to allow for passive extracorporeal arteriovenous circulation (arrows, Fig. 1). When the arterial backflow reached the predetermined backflow distance, the sampling stopcock was opened and 0.6 ml blood withdrawn into the sampling syringe. Next the transducer double-stopcock was reset and the blood within the arterial line carefully flushed back into the arterial circulation using 1±2 ml of flushing solution delivered by a syringe pump. Finally, the stopcock was cleaned with flushing solution. We recorded the following covariables: patient's age, mean arterial pressure (MAP), central venous pressure (CVP), heart rate, size of the venous and arterial catheters, total indwelling time since catheter insertion, inspiratory peak and positive end-expiratory pressure, and body temperature. Statistical analysis For each backflow distance (10 cm, 20 cm and 30 cm) and for the two paired control samples, we calculated the mean and the differences between the two samples (conventional minus backflow; first conventional versus second conventional). Likewise, the differences for the 90 duplicate analyses were determined. The differences were plotted against the respective means to obtain Bland-Altman plots [26]. We plotted two lines for reference: (a) lines denoting  2 standard deviations based on the control samples and (b) of lines indicating intralaboratory precision criteria for acceptable performance as published by the Centers of Disease Control (CDC) and the Health Care Financing Administration [27]. In further statistical analysis we checked whether the order of withdrawal (conventional first, backflow second or vice versa) had any bearing on the differences (non-parametric tests for unpaired samples, order of withdrawal serving as class variable). One-way analysis of variance was employed to compare the differences between conventional and backflow samples and the differences between the two control measurements. Post hoc tests using Scheffe's procedure were carried out in the case of significant results. Additional multivariable analyses were used to test for potential confounding by the above-listed covariates. Analyses were performed independently for each of the eight parameters. All analyses were programmed using Statistical Analysis Software (SAS Version 6.12, SAS, Cary, N. C.). The probability of a type I error of less than 0.05 was considered significant. All tests were two-tailed. With the exception of ANOVA post hoc tests, no adjustments were made for multiple testing.

Results A total of 240 blood samples were taken from 120 neonatal and pediatric intensive care patients. The median age of the patients was 9 months (range 0±15.9 years). Newborns accounted for 28 % of all patients. Table 1 summarizes further demographic data, the backflow times and mean laboratory values stratified by backflow distance or control samples. Passive backflow was achieved in all of the patients. Backflow times varied considerably between patients: the median time amounted to 20 s (range 3±35 s) for 10 cm distance, 22 s (range 5±58 s) for the 20 cm distance and 22.5 s (range 11±65 s) for the 30 cm distance.

1067

Table 1 Summary of patient characteristics, as well as the recorded backflow times and laboratory results (conventional samples), for each of the four study groups. Variables with considerable devi-

ation from the normal distribution are presented as median and interquartile range (IQR)

10 cm

20 cm

30 cm

Control

0.83 (0.1±4.0) 61.8  11.8 6.8  3.3 55.6  12.0 142.6  23.2 21.9  1.6 20.7  2.1 2 (2±4) 2 (1±4) 21.1  3.3 26/4 5.7  3.3 37.3  0.8 23.5  11.5 7.40  0.06 5.41  0.90 9.49  2.45 110.64  25.05 136.57  8.12 3.86  0.42 1.18  0.11 5.61  2.57

1.3 (0.16±5.9) 71.5+19.6 8.7+2.0 63.8+19.4 129.7  21.2 22.6  1.0 20.9  1.2 2 (1±3) 3 (1±4) 23.0  2.4 25/5 5.6  0.9 37.1  1.0 ± 7.40  0.06 5.01  0.97 12.46  4.11 106.08  24.23 137.76  5.75 3.86  0.39 1.17  0.08 6.25  2.68

Mean  standard deviation Age (years, median, IQR) Mean arterial pressure (mmHg) Central venous pressure (mmHg) Arteriovenous difference (mmHg) Heart rate (beats/min) Size of arterial cannula lumen (gauge) Size of venous catheter lumen (gauge) Time arterial cannula in place (days, median, IQR) Time venous catheter in place (days, median, IQR) Peak inspiratory pressure (cmH2O) Ventilated/spontaneous breathing PEEP (cmH2O) Core temperature (C) Duration backflow (s, median, range) pH (units) CO2 (kPa) PO2 (kPa) Hemoglobin (g/l) Sodium (mmol/l) Potassium (mmol/l) Calcium (mmol/l) Blood glucose (mmol/l)

0.9 (0.1±4.2) 58.7  13.3 8.4  3.5 50.8  12.9 129.7  29.4 22.1  2.0 21.3  1.6 3.5 (2±5) 4 (2±5) 21.6  3.2 24/6 5.3  1.4 37.4  0.7 19.9  7.9 7.42  0.05 5.53  0.84 10.73  3.72 104.76  19.15 137.03  3.95 4.4  0.42 1.12  0.07 5.6  2.30

Multivariable analysis considering potentially confounding variables revealed significant differences between the distances (adjusted mean 10 cm = 18.1 s, 20 cm = 24.9 s and 30 cm = 25.2 s). A model containing the factors flow distance, arteriovenous pressure difference and size of venous cannula explained 34 % of the observed variance in backflow times (F = 9.74, df = 4/77, p < 0.0001). No other variables (e.g. patient age) remained significant. Figures 2 presents the Bland-Altman bias plots for blood gases, electrolytes, hemoglobin and glucose. The plots show the 95 % confidence interval derived from paired conventional sampling and from the CDC suggested precision criteria [27]. Table 2 represents the mean differences (bias) and the precision (2 standard deviations of the differences) between conventional sampling and the backflow technique, and between the paired control samples. Negative differences indicate that the sample from the backflow technique yielded a higher result. For controls, negative differences indicate that the second sample/ analysis yielded a higher result. As to the comparison of backflow and conventional techniques, the difference was independent of the sampling order (all p > 0.25), except for hemoglobin and PCO2. For these parameters, the bias was slightly larger when backflow samples were obtained first, indicating that backflow values tended to be affected by the order of sampling (bias hemoglobin, backflow first sample = ±1.8 g/l versus con-

0.24 (0.05±2) 60.4  13.6 7.0  4.1 53.1  14.2 141.6  20.8 21.9  1.9 21.5  1.4 4 (2±6) 4 (2±7) 21.5  4 26/4 4.9  1.2 37.3  0.5 25.4  14.8 7.40  0.06 5.7  1.14 10.35  2.34 114.40  18.85 138.44  5.72 4.16  0.61 1.18  0.12 5.1  1.60

ventional method first sample = 1.3 g/l, p = 0.016; PCO2, ±0.15 mmol/l versus 0.04 mmol/l, p = 0.003). As shown by the Bland-Altman plot, these systematic effects on the bias are small compared to the observed precision. Differences and distribution of the biases were compared using the following five methods: conventional versus the three backflow distances, paired conventional samples and duplicate analyses from the same sample. The main findings are: the analysis of blood gases showed the same distribution of bias for all five methods in the unadjusted and in the adjusted models (all F < 1.7, df = 8/74; p > 0.11), indicating that the backflow technique provides results as reliable as those of the conventional technique. The precision for hemoglobin values showed a wider spread than that suggested as acceptable performance [27]. However, no statistical difference between the five methods was observed. The most likely explanation for the limited precision is the sedimentation occurring in the sample which was analyzed second. For electrolyte analysis there was no difference between the methods for ionized calcium or for potassium. It should, however, be noted that a few results for ionized calcium were up to 0.2 mmol/l higher with the backflow technique than with the conventional technique. Systematic differences between the conventional and the backflow methods were detected for sodium and for blood glucose. The interpretation of the data presented

1068

Fig. 2 Bland-Altman plots for the comparison of the two sampling techniques (30 cm, 20 cm, 10 cm backflow distance versus conventional sampling, each n = 30) and the control experiments. Control indicates the comparison of two conventional samples from the same patient (n = 30), duplicate indicates the results obtained from duplicate analysis of the same sample (n = 90). The thin dashed line indicates 2 standard deviations around the mean bias from the control experiment (two conventional samples from the same patient). The dashed-dotted line denotes the ªrecommended acceptable rangeº as suggested by the Centers of Disease Control (CDC) and the Health Care Financing Administration [27]

1069

Fig. 2

1070

Table 2 Bias (mean differences) and precision (2 standard deviations of the difference) of conventional minus backflow technique (10 cm, 20 cm and 30 cm) and of first minus second sample (control/duplicates); One-way ANOVA was performed for comparison

pH (units) CO2 (kPa) PO2 (kPa) Hemoglobin (g/l) Sodium (mmol/l) Potassium(mmol/l) Calcium (mmol/l) Blood glucose (mmol/l)

of the three distance groups. Negative values (bias) indicate that the backflow sample or the repeat conventional sample (control) returned a higher value

10 cm (n = 30)

20 cm (n = 30)

30 cm (n = 30)

Control (n = 30)

Duplicates (n = 90)

p

0/0.02 0.03/0.78 0.12/1.76 0.66/15.62 ±0.43/2.02 ±0.02/0.38 0.01/0.04 ±0.78/3.56

0/0.02 ±0.08/0.68 ±0.18/2.04 ±0.37/6.16 ±0.59/2.04 ±0.01/0.62 ±0.01/0.06 ±1.11/2.30

0/0.04 ±0.11/0.36 0.01/1.82 ±1.1/6.82 ±0.9/1.98 ±0.08/0.62 ±0.01/0.10 ±0.5/0.88

0/0.02 ±0.04/0.4 0.11/1.34 0.48/10.98 ±0.56/1.74 ±0.03/0.26 0/0.02 ±0.12/0.72

0/0.02 0.01/0.4 ±0.02/1.98 0.37/7.56 0.05/1.66 0/0.16 0/0.06 ±0.05/0.78

0.07 0.09 0.96 0.34 < 0.0001 0.18 0.14 < 0.0001

in Table 2 and Fig. 2 shows that the backflow technique, as well as paired conventional sampling, resulted in slightly higher sodium levels (mean difference less than 0.9 mmol/l) of the second sample. However, this difference is within the suggested range for acceptable precision [27]. Profound differences between conventional and backflow sampling were observed for blood glucose. For all backflow distances, the backflow technique returned higher values. This mean bias deviated significantly from zero (Table 2). The bias and the imprecision was higher with shorter backflow distances. Multiple regression techniques were employed to identify factors that accounted for this difference. The best model explained 30 % of the observed variance and contained the following significant factors: baseline blood glucose level (positive association, higher blood glucose levels are associated with a larger bias towards higher values from the backflow technique), size of the arterial cannula, sequence of sampling and age of patient. None of the following variables further improved the model: hemoglobin, pH, site of cannula, size of venous cannula, mean arterial blood pressure, backflow time, time since insertion of cannula, body core temperature or central venous pressure.

Discussion Blood conservation is a matter of significant concern in pediatric critical care medicine. To advance this issue at little extra expense, we applied a simple method using existing and routinely used devices for sampling from indwelling arterial lines. The crucial factor of this method is the use of the pressure transducer equipment to provide an arteriovenous passive backflow. After the arterial pressure has driven a sufficient amount of blood past the sampling stopcock, blood should no longer be contaminated by the flushing solution. Based on the 0.6 ml volume discarded during routine sampling in our

unit, we chose backflow distances of 10 cm, 20 cm and 30 cm proximal to the sampling stopcock, corresponding to withdrawn volumes of about 0.6 ml, 0.8 ml and 1.0 ml, respectively [28, 29]. The main finding of our study is that the backflow technique provides unbiased and reliable results for blood gases, electrolytes and hemoglobin. The precision of the hemoglobin measurement fell short of the range for acceptable precision suggested by the CDC [27]. The duplicate analyses showed that this imprecision was even obtained when re-analyzing the same sample. Despite much narrower variation when calibrating the analyzer with the calibration fluid provided by the manufacturer, a considerable variability of hemoglobin determination appears to be an inherent property of the method in routine use. The small systematic deviation in the sodium results (lower values obtained from conventional sampling) indicates potential dilution of the conventional sample by the flushing medium (sodium content 131 mmol/l). However, the observed bias is less than 1 mmol/l and, therefore, clinically not relevant. In contrast, blood glucose determinations by the backflow technique showed a considerable bias towards higher values. This bears the risk of missing hypoglycemic values in at-risk patients. We confirmed the validity of the determinations using the conventional sampling technique by simultaneous capillary sampling (data not shown). The mean overestimation of the true blood glucose by 0.5 mmol/l at 30 cm backflow distance and by 1.1 mmol/l at 20 cm introduces a bias that may be of clinical importance. We therefore recommend using the 30 cm backflow distance if blood glucose determinations are desired. Since the direction and magnitude of the ªerrorº are unpredictable, we would not recommend using this method if knowledge of blood glucose levels is crucial to the care of the patient. However, the majority of intensive care patients are at low risk of hypoglycemia, unless a primary metabolic or endocrine disorder is present. While the most accurate method to determine blood glucose levels should be employed in

1071

these patients, for routine monitoring it may provide satisfactory enough results to be of use. Apart from this limitation, the backflow technique offers several important advantages: first, in our unit, the method can be applied without incurring extra costs. Second, in critically ill newborns, the blood saved may become clinically relevant: for a baby weighing 3 kg with a calculated circulating blood volume of 240 ml, the savings from 5±6 arterial blood gas determinations per 24 h amount to 2.5±3 ml/day, corresponding to 1 % of the circulating blood volume or approximately the unstimulated erythropoiesis rate. Third, the method reduces manipulation of the sampling stopcock when used with closed sampling ports, one of the aims of preventive measures against nosocomial infection. Some potential caveats of the study design and of the method require consideration. An important issue is the definition of an ªacceptable rangeº of precision. With the exception of hemoglobin determinations, the suggested requirements for the intralaboratory precision published by the Centers of Disease Control were wider than the precision obtained from two subsequent samples taken from the same patient. Moreover, the subsequent determination of samples from the same patient is the clinical gold standard for management and, therefore, the most appropriate comparison. The study did not accurately compare the flushing volumes between the conventional and the backflow method. However, 1±2 ml of flushing volumes were sufficient to clear the system and it is important to note that the stopcock and

the line between sampling port and indwelling cannula must be flushed anyway, regardless of the method used. Further limitations are connected to the suggested method. First, the method is suited to arterial sampling only. Second, the equipment used in our study required the presence of central venous access to drain backflow flushing solution. Therefore backflow depends on the patency of the venous access. Backflow times can be shortened by rinsing the arterial or venous pressure line to ambient pressure or into a sampling bag. Third, we did not use the preferred sampling port for closed sampling. The main reason was that the pressure transducer kit used in our unit is provided with a conventional sampling stopcock and we aimed to minimize additional manipulation of the existing tubing system. Finally, the reason for the observed bias in glucose determinations remains elusive. In patients at risk of hypoglycemia, therefore, a precise sampling method for monitoring blood glucose levels should be sought. In conclusion, in patients with indwelling arterial catheters, iatrogenic blood loss from collecting discard volumes prior to the actual sample can be minimized by establishing an arteriovenous backflow using standard pressure transducer devices. We showed that small backflow distances between 10 cm and 30 cm provide accurate laboratory results for blood gases, electrolytes and hemoglobin. Because of overestimation of blood glucose levels, the method should not be used for monitoring blood glucose levels in patients at risk of hypoglycemia.

References 1. Smoller BR, Kruskall MS (1986) Phlebotomy for diagnostic laboratory tests in adults: pattern of use and effect on transfusion requirements. N Engl J Med 314: 1233±1235 2. Lanuza DM, Jenrich JA 1976) The amount of blood withdrawn for diagnostic tests in critically ill patients. Heart Lung 5: 933±938 3. Lin JC, Strauss RG, Kulhavy JC, et al. (2000) Phlebotomy overdraw in the neonatal intensive care nursery. Pediatrics 106: 1±6 4. Henry ML, Garner WL, Fabri PJ (1968) Iatrogenic anemia. Am J Surg 151: 362±363 5. Chernov B, Salem M, Sacey J (1991) Blood conservation ± a critical care imperative. Crit Care Med 19: 313±314 6. Smoller BR, Kruskall MS, Horowitz GL (1989) Reducing adult phlebotomy blood loss with the use of pediatricsized blood collection tubes. Am J Clin Pathol 91: 701±703

7. Wilson JR, Gaedeke MK (1996) Blood conservation in neonatal and pediatric patients. AACN Clin Issues Crit Care Nurs 2: 229±237 8. Maier RF, Obladen M, Kattner E, et al. (1998) High- versus low-dose erythropoietin in extremely low birth weight infants. The European Multicenter rhEPO Study Group. J Pediatr 132: 866±870 9. Shannon KM, Keith JF, Mentzer WC, et al. (1995) Recombinant human erythropoietin stimulates erythropoiesis and reduces erythrocyte transfusions in very low birth-weight preterm infants. Pediatrics 95: 1±8 10. Zimmermann JL, Dellinger RP (1993) Initial evaluation of a new intra-arterial blood gas system in humans. Crit Care Med 21: 495±500 11. Shapiro BA, Mahutte K, Cane RD, Gilmour IJ (1993) Clinical performance of a blood gas monitor: a prospective, multicenter trial. Crit Care Med 21: 487±494

12. Smith ML, Smith NT, Nesseler ESE (1993) In-line measurement of electrolytes, glucose and blood gases. Int Anesthesiol Clin 31: 159±180 13. Wong DK, Jordan WS (1995) Microprocessor-based near realtime bedside blood chemistry monitor. J Clin Monit Comput 9: 95±102 14. Bailey PL, McJames SW, Cluf ML, et al. (1998) Evaluation in volunteers of the VIA V-AGB automated bedside blood gas, chemistry and hematocrit monitor. J Clin Monit Comput 14: 339±346 15. Morgan C, Newell SJ, Ducker DA, et al. (1999) Continuous neonatal blood gas monitoring using a multiparameter intra-arterial sensor. Arch Dis Child 80: 93±98 16. Widness JA, Kulhavy JC, Johnson KJ, et al. (2000) Clinical performance of an in-line point-of-care monitor in neonates. Pediatrics 106: 497±504

1072

17. Petruzzi WT, Parker MA, Lichtenthal PR, et al. (1993) A clinical evaluation of a blood conservation device in medical intensive care unit patients. Crit Care Med 121: 501±506 18. Silver MJ, Li YH, Gragg LA, et al. (1993) Reduction of blood loss from diagnostic sampling in critically ill patients using a blood-conserving arterial line system. Chest 104: 1711±1715 19. Silver MJ, Jubran H, Stein S, McSweeney T, Jubran F (1993) Evaluation of a new blood-conserving arterial line system for patients in intensive care units. Crit Care Med 21: 507±511 20. Thorpe S, Thomas AN (2000) The use of a blood conservation pressure transducer system in critically ill patients. Anaesthesia 55: 27±31

21. Woda RP, Dzwonczyk R, Buyama Ch, et al. (1999) On the dynamic performance of the Abbott Safeset blood-conserving arterial line system. J Clin Monit Comput 15: 215±221 22. B Braun Medical, Berlin, Germany (1999) Haemofix ± pressure-monitoring set for closed blood sampling (catalogue) 23. pvb Medizintechnik, Kirchseeon, Germany (1994) Closed arterial blood sampling system (catalogue) 24. Rhoads MK, Kariman K (1979) A simple and safe method for monitoring of central venous and pulmonary artery pressures with a single transducer. Crit Care Med 7: 174±175 25. Calisi CM, Weigelt JA (1983) Central venous and pulmonary artery pressure measurements with a single transducer. Crit Care Med 11: 128±129

26. Bland JM, Altmann DG (1986) Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1: 307±310 27. Ehrmeyer SS, Laessig RH, Leinweber JE, Oryall JJ (1990) Medicare/CLIA final rules for proficiency testing: minimum intralaboratory performance characteristics needed to pass. Clin Chem 36: 1736±1740 28. Wallace HH (1987) Obtaining reliable plasma glucose and potassium values from intraarterial catheters. Heart Lung 1: 20±23 29. Weibley RE, Riggs CD (1989) Evaluation of an improved sampling method for blood gas analysis from indwelling arterial catheters. Crit Care Med 17: 803±805