| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Address correspondence to Christine N. Papadea, Ph.D., Department of Pathology and Laboratory Medicine, Medical University of South Carolina, PO Box 250908, Charleston, SC 29425, USA; tel 843 792 1189; fax 843 792 4811; e-mail papadeacn{at}musc.edu
 |
Abstract |
|---|
 Top Abstract IntroductionMethod and MaterialsResultsDiscussionAcknowledgementsReferences |
|---|
(received 1 January 2002; accepted 4 February 2002)
Keywords: point-of-care testing, neonatal/pediatric intensive care, clinical chemistry assays
|
Introduction |
|---|
|
TopAbstractIntroductionMethod and MaterialsResultsDiscussionAcknowledgementsReferences |
|---|
One of our major concerns as health-care providers and clinical laboratorians is the validation of measurements by POC devices compared to the primary instruments and methods used in the central clinical laboratory, to ensure the transferability and consistent interpretation of test results.
In this report we describe precision, linearity, and comparison studies of i-STAT measurements of blood sodium (Na), potassium (K), chloride (Cl), ionized calcium (iCa), glucose, urea nitrogen (UN), creatinine, hematocrit (hct), pH, partial pressure of oxygen (pO2), and partial pressure of carbon dioxide (pCO2) performed at the MUSC Medical Center. Our aim was to determine if results obtained by this analyzer, under consideration for bedside use in neonatal and pediatric intensive care units, including extracorporeal membrane oxygenation (ECMO) settings, were comparable to those obtained by the clinical laboratory’s primary methods.
Accreditation regulations require clinical laboratories to verify or establish performance criteria for reproducibility (precision), linearity (analytical range), and accuracy before reporting patient test results by a newly introduced analytical method. Using aqueous commercial solutions, blood products enriched in vitro, and patient blood samples, we performed the required and additional studies to comply with accreditation regulations and attain our specific objective. We also report here the discrepancies that we found for blood creatinine assays after observations in our neonatal intensive care unit detected an i-STAT measurement problem that would have affected clinical management.
|
Method and Materials |
|---|
|
TopAbstractIntroductionMethod and MaterialsResultsDiscussionAcknowledgementsReferences |
|---|
): the hand-held analyzer, and the single-use cartridges that are available in different configurations of measured analytes, including Na, K, Cl, iCa, hct, glucose, UN, creatinine, pH, pO2 and pCO2. Calculated levels of hemoglobin, total CO2, bicarbonate, extracellular base excess, and anion gap are derived from the relevant measured constituents by the analyzer software. The cartridges contain the components needed to perform the individual analyses, including miniaturized ion-specific or enzyme-coated electrodes, a single point calibrant, and a waste reservoir. Whole blood, 65 to 100 µL (depending on the number of electrodes in the cartridge), is introduced by syringe or capillary tube into a port in the cartridge. The cartridge is immediately inserted into the hand-held analyzer. A motor in the analyzer moves the calibrant across the electrodes to perform a single-point calibration, and then pushes the test sample forward for contact with the electrodes. The test cycle is completed in 120 sec and the results are displayed . If the analyzer detects electronic or mechanical errors, the result for the affected substance is suppressed, being replaced by an error code.
|
Primary analyzers. The clinical laboratory instruments included the ABL 725 blood gas analyzer (Radiometer, Copenhagen, Denmark), Vitros 750 chemistry analyzer (Ortho-Clinical Diagnostics, Rochester, NY), and STKS hematology analyzer (Coulter Corporation, Miami, FL). The Radiometer measurements in whole blood are based on potentiometry (pH, pCO2, and electrolytes) and amperometry (pO2 and glucose) [11], the Vitros 750 chemistry analyzer uses dry multilayered chemistry slides and reflectance photometry [12] for plasma UN and creatinine measurements, and the Coulter STKS electronic counting of blood cells is based on conductometry [11]. These instruments and the respective methods, the main analytical systems for many years in the MUSC clinical laboratory, had been validated repeatedly for linearity, calibration, and reproducibility. Having well-established performance characteristics, these test methods were used for comparison with the PCA results.
Method comparison studies. Patient blood samples were separated into 2 age-related subgroups for comparison of whole blood Na, K, Cl, iCa, hct, pH, pO2, pCO2, and glucose; blood UN and creatinine was determined in only one subgroup. Samples from patients <3 mo of age in the neonatal intensive care unit were collected in sterile arterial blood gas syringes (1 ml; 50 units lyophilized lithium heparin, Owens-Brigam Medical Co, Morganton, NC) and samples from patients >3 mo of age were drawn from heparinized arterial lines into sterile, 3 ml plastic syringes. Blood samples were analyzed first by the Radiometer ABL to complete physician-ordered laboratory tests and then by the i-STAT cartridges. Due to the limited volume of most blood samples, analyses were performed by single testing, except as stated. The number of samples tested varied among the three cartridge types, reflecting the ages of the patients and the availability of samples. Thus, 19 to 75 samples were tested by the EC8+ cartridge, 26 to 58 samples were tested by the EC6+ cartridge, and 53 samples were tested by the G3+ cartridge. Seventy-five samples were tested for hct by the EC 6+ cartridge and the Coulter STKS analyzer. Forty-five blood samples from patients >3 mo of age were tested for UN (EC8+ cartridge) and creatinine, and then centrifuged to obtain plasma for UN and creatinine measurements by the Vitros 750 analyzer. Following the clinical observation of creatinine discrepancies, 59 samples from patients >3 mo of age were tested in the second creatinine method comparison study.
Reproducibility (precision) studies. Between-day precision for the analytes, except hct, was determined using 3 concentrations of one lot of aqueous controls purchased from i-STAT Corp. The aqueous solutions were tested once daily in duplicate on 13 or 14 days. Each sample was tested by two cartridges of each type; two lots of each cartridge type were used. Hct reproducibility was determined by testing patient samples in duplicate. To assess the ability of i-STAT cartridges to duplicate analyte measurements in blood samples, 10 patient blood samples received in heparinized syringes by the clinical laboratory for testing unrelated to this study were used. The samples were obtained over a 5-day period when a sufficient surplus volume was available following the completion of physician-ordered testing.
Linearity. The linearity performance of the PCA for analytes on the EC8+, EC6+, G3+, and creatinine cartridges was determined using aqueous solutions from i-STAT or RNA Medical (Acton, MA). Linearity studies for Na, K, pH, pCO2, and hct were performed on only one cartridge type. Five calibration solutions were used for Na, K, Cl, iCa, UN, creatinine, pH, pO2, and pCO2 and 6 were used for glucose. Linearity for hct measurements was determined with a set of samples prepared by mixing increasing amounts of red blood cells with a constant amount of compatible plasma. All solutions and prepared samples were tested in duplicate.
Comparison methods. Patient samples tested by the PCA were assayed by the primary laboratory methods to determine the concordance with the primary results. Whole blood measurements for Na, K, Cl, glucose, iCa, pCO2, and pO2 were compared with those by the Radiometer ABL analyzer; whole blood UN and creatinine results were compared with plasma results by the Vitros 750 analyzer; hct measurements were compared with those determined by the Coulter STKS analyzer. The data were interpreted according to standard criteria [13].
Inter-instrument reproducibility. Inter-instrument reproducibility is routinely assessed periodically in our clinical chemistry laboratory for quality assurance purposes. Before beginning the comparison studies for the i-STAT cartridges, we checked the inter-instrument reproducibility of 3 Radiometer blood analyzers for Na, K, Cl, iCa, glucose, hct, pH, pCO2, and pO2. Ten patient blood samples received in heparinized syringes for testing by physician-orders were analyzed once by each analyzer to obtain 3 results for each sample.
Extracorporeal membrane oxygenation (ECMO). To assess the feasibility of the using i-STAT cartridges in settings at the ECMO bedside, a neonatal ECMO circuit primed with blood was set up in the clinical laboratory by an ECMO specialist registered nurse. The ECMO circuit provided an in vitro system which allowed adjustment of pO2, pCO2, and calcium over a wide range of concentrations. Patient samples were not used for these studies. One unit each of red blood cells and compatible plasma preserved with citrate phosphate dextrose were combined in appropriate proportions to obtain a mixture comparable to normal whole blood for priming the circuit. Ultra-pure nitrogen, oxygen, and carbon dioxide were used for gas flow (sweep gas) through the membrane oxygenator.
Calcium gluconate (10 g/dl), a solution typically added with citrated blood for ECMO, was used to enrich the citrate-reduced calcium concentration in the circuit. Blood was recirculated and each sweep gas or the calcium gluconate solution was introduced separately into the circuit at the oxygenator port. The ECMO specialist gradually altered the partial pressures of CO2 and O2 and the concentration of calcium to reach the levels of interest (pCO2, 15 to 130 mm Hg; pO2, 20 to 500 mm Hg; and iCa 0.25 to 3.0 mmol/L) in the circuit blood. As the sweep gases were adjusted or the calcium gluconate was added, blood samples were drawn from the circuit into plain syringes at intervals of 2 to 10 min. A laboratory technologist analyzed the samples immediately by Radiometer and PCA instruments.
In separate experiments, blood samples were prepared over a wide range of K concentrations simulating those expected in ECMO samples. By adding increasing amounts of red cell hemolysate or red blood cells to a constant volume of plasma, samples with K as high as 9 mmol/L and hematocrits as high as 70% were created for analysis by the PCA, the Radiometer, or the Coulter STKS instruments.
Statistical analysis. Descriptive analysis was used to estimate precision (Statview software, Abacus Concepts, Inc., Berkeley, CA). Deming linear regression analysis in EP_Suite for Windows software (MarChem Associates, Concord, MA) was used for the linearity and the method comparison data.
|
Results |
|---|
|
TopAbstractIntroductionMethod and MaterialsResultsDiscussionAcknowledgementsReferences |
|---|
). Measurements for pCO2, pO2, and creatinine were less precise (CVs 2–6.5%), but were within the manufacturer’s reproducibility limits (CVs 
10% ) for these analytes in aqueous solutions. The %CVs measured by the i-STAT compared favorably with those in aqueous quality controls used in our clinical laboratory, except creatinine, which has a CV 3% at 1.0 mg/dl by the Vitros 750 method. Hct reproducibility was assessed in blood samples, as described below.
|
2.5%) among 3 cartridge types for duplicate measurements of 10 patient samples, with the exception of hct (Table 2). In addition, good precision for pH, pCO2, and pO2 (CVs <1, <4, and <6%, respectively) was found over a wide range of values for each of these analytes. The reproducibility of these duplicate measurements compared favorably with the triplicates of a different group of samples by the Radiometer blood gas analyzers (Table 3).
|
|
). In contrast, the Radiometer measurements for hct (Table 3) and plasma creatinine assays in duplicate by the Vitros 750 for each of 10 patient samples (CV <9.5%, data not shown) were more reproducible. These findings may be due, in part, to different principles of measurement or different sample types: the Radiometer hct is calculated from the directly-measured hemoglobin, the i-STAT hct is determined by conductivity, the Vitros plasma creatinine is measured as a colored end-product of enzymatic reactions, and the i-STAT measures the creatinine reaction end-product, hydrogen peroxide, by amperometry.
Linearity performance refers to the verification of the analytical range over which the measurements can be made without modifying samples by dilution [13]. The linearity data for each analyte were reduced by regression analysis to obtain the coefficients for the slope, y-intercept, standard error (SE; Sy/x), and correlation. The derived coefficients (Table 4) were acceptably close to ideal for slope (1.0), y-intercept (0), SE (0), and correlation (1.0) and they agreed reasonably with the analytical ranges published by the i-STAT Corporation.
|
The ranges found for blood Na, UN, creatinine, pH, and iCa were similar to the i-STAT performance specification ranges; those for blood Cl and glucose exceeded the published ranges, but values as wide as those specified by the manufacturer for K, pO2, and pCO2 were not attained. Results of additional studies designed to test the analytical limits for hct, K, iCa, pO2, and pCO2 are discussed below.
The ability to measure K, hct, iCa, pO2, and pCO2 in a wide range of concentrations was clinically important to the ECMO team interested in using the i-STAT at the bedside during ECMO procedures. By adjusting the concentrations of K and red blood cells in the series of prepared samples and the levels of iCa, pO2, and pCO2 to simulate the conditions in the ECMO circuit, we determined that the highest limits measured by the i-STAT and verified by the Radiometer ABL or the Coulter STKS were: K, 7.9 mmol/L; hct, 69%; iCa, 2.6 mmol/L; pO2, 394 mm Hg; and pCO2, 89 mm Hg (Table 4). Thus, the range of pO2 could be measured ~100 mm Hg higher than had been determined with aqueous solutions, but no other gains with enriched blood samples were realized.
The ranges of analyte concentrations in the patient samples separated into 2 age-related subgroups were widely distributed, included normal and abnormal values, and spanned similar ranges in both the primary and the test methods. Table 5 summarizes the Deming regression analyses, which were based on the primary method measurements (the independent [x] variables) compared with the test values (dependent [y] variables). The Deming method takes into account the imprecision of measurements by both methods and computes the coefficients for the slope (estimate of proportional bias), the SE (variance around the slope), the y-intercept (estimate of constant bias), and the correlation coefficient (r), which indicates the linear association between the x and y variables.
|
). In contrast, highly correlated results were revealed for K, iCa, pH, pCO2, and UN with near-ideal slope coefficients (between 1.002 and 1.118), negligible y-intercepts, and r-coefficients above 0.92. The pO2 values by the i-STAT were biased low, probably indicating the inherent instability of the O2 in the samples during transfer between the two analyzers.
Following completion of the initial evaluation, PCA was implemented for clinical use. Within 1 mo, the PCA creatinine assay was discontinued after the clinical staff observed analytical discrepancies for creatinine concentrations in blood from several babies in the neonatal intensive care unit. A second creatinine method comparison was performed with 59 patient samples and creatinine cartridges in a production lot that was different from the lots previously evaluated. The regression coefficients of the initial and second creatinine comparison studies (Table 5), the scatterplots, and individual differences plots of the data (Fig. 2) show that the agreement between the Vitros (plasma) and the PCA measurements in the initial study was better than in the second study.
|
0.3 mg/dl. Discrepant results between the two studies could be due, in part, to different distributions of creatinine concentrations in the samples, different numbers of samples, and different production lots of the creatinine cartridges. |
Discussion |
|---|
|
TopAbstractIntroductionMethod and MaterialsResultsDiscussionAcknowledgementsReferences |
|---|
While many reports of the i-STAT system have described studies with various cartridges including the G3+, the EC6+, and others in settings as diverse as a hemodialysis unit [4], a helicopter [5], and an emergency department [9], we have found no published evaluation of the creatinine cartridge or of the use of the PCA in ECMO settings. Several studies involved only one cartridge and fewer blood samples [1–3,5,6,8] than we used, or compared a few analytes in whole blood to those in plasma [1,3–5,9]. One study design used two cartridges for 4 analytes: blood gases, pH, and iCa- [10], and others used 3 cartridge configurations for evaluation of multiple analytes in small [4] or large [7] sample numbers, similar to the protocol we used.
Our analytical results with the EC6+ cartridge (Na, K, and iCa )and the G3+ cartridges (pH, pCO2, and pO2) agreed with the findings of Murthy et al [7] for application of the PCA in neonatal and intensive care settings. The earlier report presented results with another cartridge, the 6+, in addition to the EC6+ for Na and K, whereas we evaluated the large panel EC8+ and the single analyte creatinine cartridges. Despite differences in experimental design among the various i-STAT studies, our results agree in general with those in previous reports, except one [10]. We found no discrepancies for pCO2 as reported by Ng et al [10], a success we attribute to the fact that the brand of arterial line draw syringes identified in the earlier study has not been used in our hospital.
Our expectation for PCA creatinine measurements at the bedside was short-lived. We stopped using the creatinine cartridges when unexpectedly high creatinine levels in neonatal blood samples were not confirmed by the clinical laboratory results. In the second method comparison study, we verified the clinical observations: measurements were higher by the PCA than by the routine laboratory method. While a repeat linearity study was not performed, the differences in the comparison data suggest the PCA creatinine may not be optimized for measuring concentrations as low as 0.3 mg/dl, a level that is not unusual in babies < 1 mo of age, depending on gestational age, birthweight, and other factors [14]. We provided the data of our initial and repeat studies to an i-STAT technical representative who replied by telephone that our results had been reproduced by the manufacturer, a finding that pointed to a manufacturing problem. Our request for written additional information has not been answered and we have not reinstated PCA creatinine assays.
The role of the ECMO experiment, unique among i-STAT reported studies, provided an in vitro system by which the iCa, pO2, and pCO2 analytical ranges could be tested for the PCA for its utilization with ECMO samples. Additional benefits afforded by this experiment included: sampling from patients was avoided, analyte concentrations in blood were easily prepared across wider ranges than would be be feasible in patients, the blood volume per sample was not a limiting factor, and the work was readily and rapidly completed.
Our linearity studies on all cartridges using commercially available materials or supplemented blood samples provided information needed before this system could be made available for neonatal and pediatric intensive care as well as for ECMO critical care. Our work using commercial aqueous solutions confirmed the analytical ranges claimed by i-STAT for Na, UN, creatinine, iCa, and pH, but did not verify those claimed for K, Cl, glucose, hct, pO2 and pCO2. These discrepancies may have been due to the lot of aqueous solutions, the cartridge lots, or unknown factors. These observations underscore the responsibility of laboratory users to verify the analytical claims published by manufacturers of laboratory diagnostic products and to repeat such studies periodically in order to assess between-lot variability of the products.
The reproducibility of duplicate measurements of 10 patient blood samples and that of between-day measurements of stabilized aqueous products confirmed the reports of others [4,7,9]. Our findings also suggested that the reproducibility by the PCA was comparable to that by the primary methods for all analytes except hct and creatinine. The lowest level of aqueous control for creatinine (1 mg/dl) and lowest concentration (~0.6 mg/dl) among the 10 patient samples available for duplicate measurements did not alert us to question the lowest level of sensitivity of the creatinine cartridge. A key element in conducting method evaluations is the careful selection of patient samples to obtain a sufficient number distributed according to age, clinical conditions, and analyte concentrations likely to occur in the population to be tested.
The initial method comparison studies were separated into 2 age-related patient groups with as many samples and widely distributed concentrations as possible in each subgroup. Knowing that the i-STAT would be used initially in the MUSC Medical Center neonatal and pediatric intensive care units, we considered the possibility that fetal hemoglobin (Hb F), inherently elevated in blood of patients <3 mo of age, might affect the results. Elevated Hb F was a significant cause of discordant co-oximetry measurements in our evaluations of blood gas analyzers several years ago. No apparent differences were detected for the results compared between the 2 age groups and for 3 cartridges. Our results agreed with the results of an earlier study [7] that used a single group of >100 samples from patients with ages from 1 day to 12 yr.
Based on the multiple regression analyses with various cartridges, sample numbers, and different age groups, we considered the agreement between the i-STAT and primary methods to be satisfactory for Na, K, UN, pH, glucose, iCa, and even pO2, pCO2, considering the instability of blood gases in vitro. Lower results for hct, an analyte that appears to have inconsistent assessments [1,2,9] and assay imprecision by the PCA method, led to our clinical decision to use the hct measurement only for trending and not for decisions about transfusions. Differences between results by the primary and the PCA methods may be due to the reference method that i-STAT uses for setting calibration points, to unknown substances in the samples that may cause interference [4,15], or to the different principles on which the methods are based.
The i-STAT PCA evaluation data were reviewed by a team of physicians, nurse managers, and clinical laboratory staff (technical and administrative), and the system was placed in service in early 2001, after extensive training of the neonatal and pediatric intensive care nurses who perform the bedside testing. The MUSC Clinical Laboratory Services is responsible for controlling, distributing, and monitoring cartridge utilization. The quality control program includes testing new lots of cartridges with i-STAT aqueous controls and checking components of the analyzers with the electronic cartridge simulator at scheduled intervals. On average, 2600 cartridges/mo (total for 3 configurations) are used.
The average wastage rate, attributed to cartridge, operator, and analyzer errors, has been constant at approximately 6% per month for a 10-mo period. This error rate is slightly higher than the 11 results for various analytes among all assays in 225 samples that was noted during evaluation studies performed by a few operators (experienced laboratory technologists) for a shorter period of time.
In our facility, 2 technical employees devote 30% to 80% of their full-time work to i-STAT surveillance. Additional responsibilities include initial and periodic training of the clinical staff that perform the testing, competency assessment, maintaining inventory records, and problem-solving of PCA errors. Through an information management system that interfaces analyzer units at bedside locations to a central data station in the clinical laboratory, a laboratory technologist reviews patient results, data entry, and other codes, and facilitates corrective actions based on these and other electronic reports. An additional advantage of the electronic central data station is the link it provides between the PCA and the central laboratory information system for patient records, billing, and archiving.
Infrequent spurious results (observations that exceed the majority of results known from clinical assessment or the previous results of the same subject’s samples) that had been noted occasionally in our evaluation studies occur, as well, in the clinical setting. Special precautions for collecting infant patients’ samples must be observed to prevent invalid results from pre-analytical sources. For example, samples for iCa should be collected in syringes or capillary tubes with the appropriate amount and type of anticoagulant. Using proper techniques for collecting blood from heelsticks is essential to minimize hemolysis or excess tissue fluid in the sample. When results at the bedside setting are inconsistent with the clinical impression, physicians at our hospital have an option of repeating the test with another i-STAT cartridge, but they usually order it to be performed in the clinical laboratory.
Efficiency, ease of use, small sample requirement, and generally reliable results obtained with the i-STAT PCA, its extensive evaluation, and the ongoing team effort of clinical laboratorians and health care providers have contributed to successful implementation of the i-STAT PCA in the neonatal and pediatric intensive care units of the MUSC Medical Center Children’s Hospital.
|
Acknowledgements |
|---|
|
TopAbstractIntroductionMethod and MaterialsResultsDiscussionAcknowledgementsReferences |
|---|
|
References |
|---|
|
TopAbstractIntroductionMethod and MaterialsResultsDiscussionAcknowledgementsReferences |
|---|
This article has been cited by other articles:
|
|
 |
|
  S. M. Hopfer, F. L. Nadeau, M. Sundra, and G. S. Makowski Effect of Protein on Hemoglobin and Hematocrit Assays with a Conductivity-Based Point-of-Care Testing Device: Comparison with Optical Methods Ann. Clin. Lab. Sci., January 1, 2004; 34(1): 75 - 82. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
