Original scientific paper:
Massimo Daves1*, Roberto Cemin2, Bruno Fattor3, Giovanni Cosio1, Gian Luca Salvagno4, Francesco Rizza1, Giuseppe Lippi5. Evaluation of hematocrit bias on blood glucose measurement with six different portable glucose meters. Biochemia Medica 2011;21(3):306-11. http://dx.doi.org/10.11613/BM.2011.041
1Clinical Biochemical Laboratory, Regional Hospital of Bolzano, Bolzano, Italy
2 Cardiology Division, Regional Hospital of Bolzano, Bolzano, Italy
3 Internal Medicine Division, Regional Hospital of Bolzano, Bolzano, Italy
4 Sezione di Chimica Clinica, Università degli Studi di Verona, Verona, Italy
5 U.O. Diagnostica Ematochimica, Azienda Ospedaliero-Universitaria di Parma, Parma, Italy
*Corresponding author: massimo [dot] daves [at] asbz [dot] it
Introduction: Measurement and monitoring of blood glucose levels in hospitalized patients with portable glucose meters (PGMs) is performed widely and is an essential part of diabetes monitoring, despite the increasing evidence of several interferences which can negatively bias the accuracy of measurements. The purpose of this study was to evaluate the effect of the hematocrit on the analytical performances of different PGMs as compared with a reference laboratory assay.
Materials and methods:The effect of various hematocrit values (~0.20, ~0.45 and ~0.63 L/L) were assessed in three whole blood specimens with different glucose concentration (~1.1, ~13.3, and ~25 mmol/L) by using six different commercial PGMs. The identical samples were also tested with the laboratory reference assay (i.e., hexokinase). The percentage difference from the laboratory assay (%Diff) was calculated as follows: % Diff = average PGM value - value from laboratory assay x 100 / value from laboratory assay.
Results: The %Diff of the six different PGMs were rather broad, and comprised between 56.5% and -34.8% in the sample with low glucose concentration (č1.1 mmol/L), between 40% and -32% in the sample with high glucose concentration (~13.3 mmol/L), and between –50% and 15% in the sample with very high glucose concentration (~25 mmol/L), respectively. It is also noteworthy that a very high hematocrit value (up to 0.63 L/L) generated a remarkable negative bias in blood glucose (-35%) as measured with the laboratory assay, when compared with the reference sample (hematocrit 0.45 L/L).
Conclusion: The results of this analytical evaluation clearly confirm that hematocrit produces a strong and almost unpredictable bias on PGMs performances, which is mainly dependent on the different type of devices. As such, the healthcare staff and the patients must be aware of this limitation, especially in the presence of extreme hematocrit levels, when plasma glucose assessment with the reference laboratory technique might be advisable.
Key words: portable glucose meters; hematocrit bias; analytic performance
The chronic hyperglycaemia of diabetes mellitus is associated with long-term organ dysfunction and increased risk of complications such as cardiovascular disease, renal failure, retinal and neurological diseases (1). Regular monitoring of blood glucose along with appropriate pharmacological treatment are effective to improve the glycaemic control and thereby decrease the burden of long term complications of hyperglycaemia (2). In agreement with the current position of the American Diabetes Association (ADA), self-monitoring of blood glucose (SMBG) is recommended for all patient undergoing insulin therapy (3). In particular, ADA suggests that SMBG should be used in patients on intensive insulin therapy, in patients not in pharmacological treatment but in diet therapy alone to achieve an optimal glycaemia control, and to achieve the optimal postprandial glycaemic target (4). The National Academy of Clinical Biochemistry (NACB) also recommends that SMBG should be made available to all diabetic patients on insulin therapy (5). Portable glucose meters (PGMs) are widely used by patients for home testing, and despite these devices are fast and easy to use, the patients must be trained on the correct handling. Moreover, PGMs are currently used in other clinical settings, including departments of acute and chronic care (hospital, clinics) as well as internal medical wards to monitor hypoglycaemic therapy in diabetics and patients with acute myocardial infarction. In internal medicine wards, PGMs are used directly by the healthcare staff, which should be appropriately trained to their use and maintenance, so that the risk of improper application can be eliminated or limited at least. Nevertheless, some variables can affect the efficacy of glucose monitoring by PGM also in this setting, including the hematocrit level, hypoxemia, hypotension, hypertriglyceridemia, temperature and humidity of the environment (6). The purpose of this study was to evaluate the interference of the hematocrit on the analytical performances of different PGMs as compared with the laboratory measurement.
Material and methods
To evaluate the effect of hematocrit on reliability of PGM test result, three different hematocrit levels (0.22, 0.45 and 0.62 L/L) were studied. Three levels of glucose (target ranges as č20 mg/dL, č240 mg/dL and č450 mg/dL; i.e., č1.1 mmol/L; č13.3 mmol/L and č25 mmol/L) were assessed. We obtained discarded venous blood collected in heparin tubes without separator gel (Venosafe, Terumo Europe, Leuven, Belgium) from healthy donors. Assuming that samples with very low glucose concentrations are not easily available, these were obtained by collecting the blood the day before the evaluation and maintained the whole anticoagulated blood on a rocker to allow consumption of glucose. We ensured the blood was saturated with oxygen by opening the tubes. Afterwards, a glucose stock solution (Concentration 20 g/dL) supplied by Nova Biomedical Corporation (Waltham, CA, USA)(i.e., the addition of 7.5 μL of glucose spiking solution to a 1 mL whole blood sample increases the glucose concentration by 8.3 mmol/L) was spiked into each blood collection tube to obtain the target glucose ranges andthe blood tubes were then placed on blood tubes rocker for 10 minutes to allow adequate mixing.
The initial hematocrit of the stock blood (i.e., 0.469 L/L) was assessed on a hematological analyzer (Coulter LH750 Analyser, Beckman Coulter, Fullerton, CA, USA). In our laboratory the reference interval of hematocrit values are 0.37-0.47 L/L for females and 0.42-0.52 L/L for males, respectively. Three aliquots of 1 mL samples were thereby prepared (labelled as A1, A2, A3) by adding fixed amount of packed red blood cells and plasma to achieve final hematocrit concentrations in the tubes as follows: A1: 0.22; A2: 0.45; A3: 0.62 L/L. The tubes were then placed on tube rocker for 10 minutes and glucose was tested afterwards with six PGMs and, after centrifugation of the samples (then minutes at 3500 rpm), by the reference laboratory assay (i.e., hexokinase, Olympus AU 2700, Beckman Coulter, Fullerton, CA, USA). Table 1 shows the main characteristics of the PGMs used in this study (Accu-Chek Compact Plus Roche Diagnostics GmbH, Mannheim, Germany; Breeze2 Bayer, Basel, Switzerland; One touch Vita, Life Scan Inc., Milpitas, CA, USA; Optium Xceed Abbott Diabetes Care, Oxon, UK; Ratisbonne BGM, Acon Laboratories Inc. San Diego, CA, USA; Stat-Strip Xpress Nova Biomedical, Waltham,CA, USA). Only one test strip lot was used for each PGMs. The PGMs and the laboratory assay were all calibrated according to manufacturer’s instruction. All the measurements were performed simultaneously (within 10 min) in duplicate by two skilled laboratory technologists.
Table 1. Characteristic of the GPMs used in this study.
The % difference from laboratory method (%Diff) was calculated by the average glucose measurement of each duplicate obtained from each PGM from the value obtained by the laboratory method (i.e., % Diff = average PGM value - value from the reference laboratory assay x 100 / value the reference laboratory assay).
The results of this investigation are shown in figures 1, 2 and 3 and in tables 2, 3 and 4. At low glucose concentration (~1.1 mmol/L) the %Diff from the value obtained by the laboratory method are comprised between 56.5% and -34.8%. Among the different PGMs, the Stat-Strip Xpress shows the best performance in comparison with the reference laboratory assay (%Diff between -4.3 and +8.7). It is however noteworthy that the bias of some PGMs (as compared with the reference laboratory assay) at normal hematocrit level and very low glucose concentration was broad and clinically meaningful (Figure 1).
Figure 1. %Diff from laboratory method at different hematocrit levels (0.22, 0.45 and 0.63 L/L) in the sample with very low glucose concentration (~1.1 mmol/L).
At high glucose concentration (~13.3 mmol/L) the %Diff were comprised between -32% and +40%. In such case, the modest performance by PGMs is conceivably attributable to the effect of high levels of hematocrit. Interestingly, the observed bias of PGMs was mostly negative (i.e., from -32% to -17%), with the only exception of the Ratisbonne (+40%). The Stat-Strip Xpress and the Accu-Chek showed the best performance in comparison with the laboratory assay (Figure 2).
Figure 2. %Diff from laboratory method at different hematocrit levels (0.22, 0.45 and 0.63 L/L) in the sample with high glucose concentration (~13.3 mmol/L).
At very high glucose concentration (~25 mmol/L), the %Diff was comprised between -50% and 15%. (Figure 3). As expected, the great difference from the reference laboratory assay was again attributable to the influence of extremely high hematocrit level. Even more interestingly, the sample with hematocrit value up to 0.63 also produced a remarkable bias using the laboratory technique, since the sample with a theoretically glucose concentration of 25 mmol/L yielded instead a final value of 16 mmol/L (i.e., -35%).
Figure 3. %Diff from laboratory method at different hematocrit levels (0.22, 0.45 and 0.63 L/L) in the sample with very high glucose concentration (~25 mmol/L).
Table 2. Glucose measurement (mean of two replicates) and %Diff on sample with low glucose concentration. Value from laboratory method 1.28 mmol/L (as mean of two replicate in the sample with hematocrit 0.45 L/L).
Table 3. Glucose measurement (mean of two replicates) and %Diff on sample with high glucose concentration. Value from laboratory method 13.4 mmol/L (as mean of two replicate in the sample with hematocrit = 0.45 L/L.
Table 4. Glucose measurement (mean of two replicates) and %Diff on sample with very high glucose concentration. Value from laboratory method 25.2 mmol/L (as mean of two replicate in the sample with hematocrit = 0.45 L/L.
Preanalytical variability and analytical quality both have a strong influence on the reliability of laboratory testing, on common laboratory assays (7-11) and point of care testing (POCT) (12).
The revised Clinical and Laboratory Standards Institute (CLSI, the former National Committee for Clinical Laboratory Standards) guidelines (9), and further adopted by the International Organization for Standardization (ISO), recommend that less than 5% of the samples should have a bias of ± 0.83 mmol/L (for glucose concentrations < 4.2 mmolL), or within ± 20% (for glucose concentrations > 4.2 mmol/L), when compared with the reference laboratory assay. Therefore the results of this analytical investigation confirm that the hematocrit value can substantially bias glucose measurements on PGMs and reference laboratory assay, producing a bias that largely exceed the quality requirements established by the CLSI and which can thereby be considered clinically significant. Noteworthy, the hematocrit bias appeared extremely heterogeneous and mostly unpredictable among the different PGMs, with differences as high as 40%. This finding has substantial clinical implication for both the clinical decision making and the therapeutic management, especially in subjects who are characterized by unusually high or low hematocrit values, such as newborns and critically ill patients, respectively. In particular, we have shown that this bias is also evident in samples with very low glucose levels (i.e., č1.1 mmol/L) as compared with the reference laboratory technique (Figure 1). Recently Roth-Kleiner report that before daily use in the newborn population, careful clinical evaluation of each new POCT system for glucose measurement is of utmost importance, concluding that the bench analyzer ABL 735 was the most accurate system, being however characterized by an important drawback (i.e., the blood volume needed is more than 15 times higher than for handheld PGMs) (13). So, the users of PGMs (both the healthcare personnel and the patients) must be aware of this limitation and, as recommended by Tang et al. (14), we suggest that clinicians must interpret with great caution the results obtained with PGMs in patients with abnormal (especially very high or very low) hematocrit levels. In these circumstances, the measurement of plasma glucose using the laboratory technique might be advisable.
Although the mechanism underlying the preanalytical interference of hematocrit has not been fully established, it has however been suggested that the presence of an increased number of red blood cells might (mechanically) prevent the diffusion of plasma through the layers of the test strips, thereby decreasing the volume of plasma available for the enzymatic reaction (6). This interference has been eliminated in some PGMs that measure hematocrit along with glucose in the drop of blood, operating a further correction of test results (15). Some authors have also developed a simple mathematical correction formula for some commonly used PGMs used in the United States that is effective to reduce, as claimed by the authors, the inaccuracy caused by anemia (16,17). More recently, Hoedemaekers et al. failed to observe an effect of hematocrit on the accuracy of PGMs performance in critically ill patients (18).
We would like to thank the technical staff (Maria Gabriella Groppo, Nadia Trevisan and Sara Negrisolo) for their collaborative work.
Potential conflict of interest
1. Bloomgarden ZT. Diabetes Care 2004;27:1506-14.
2. Standards of medical Care in Diabetes. Diabetes Care 2006;29(suppl):S4-42.
3. Goldstein DE, Little RR, Lorenz RA, Malone JL, Nathan DM, Peterson CM. American Diabetes Association. Test of glycemia in diabetes. Diabetes Care 2004;27(Suppl.1):S91-3.
4. American Diabetes Association. Standards of medical care in diabetes-2008. Diabetes care 2008;31(Suppl 1):S12-54.
5. Sacks DB, Arnold M, Bakris GL, Burns DE, Horvath AR, Kirkman MS, et al. Guidelines and recommendations for laboratory analysis in the diagnosis and management of diabetes mellitus. Draft version 1107. Available at: http://www.aacc.org/sitecollectiondocuments/nacb/lmpg/diabetes/1107_diab.... Accesed: April 10, 2011.
6. Montagnana M, Caputo M, Giavarina D, Lippi G. Overview on self-monitoring of blood glucose (review). Clin Chim Acta 2009;402:7-13.
7. Lippi G, Plebani M, Simundic AM. Quality in laboratory diagnostics: from theory to practice. Biochem Med 2010;20:126-30.
8. Lippi G, Simundic AM, Mattiuzzi C. Overview on patient safety in healthcare and laboratory diagnostics. Biochem Med 2010;20:131-43.
8. Müller MM. Quality and diagnostic perspectives in laboratory diagnostics. Biochem Med 2010;20:144-6.
10. Sonntag O. Quality in the analytical phase. Biochem Med 2010;20:147-53.
11. Simundic AM, Topic E, Nikolac N, Lippi G. Hemolysis detection and management of hemolysed specimens. Biochem Med 2010;20:154-9.
12. Giavarina D, Villani A, Caputo M. Quality in point of care testing. Biochem Med 2010;20:200-6.
13. Roth-Kleiner M, Stadelmann Diaw C, Urfer J, Ruffiex C, Werner D. Evaluation of different POCT devices for glucose measurement in a clinical neonatal setting. Eur J Pediatr 2010;169:1387-95.
14. Tang Z, Lee JH; Louie RF, Kost GJ. Effects of different hematocrit levels on glucose measurements with handheld meters for Point-of-Care testing. Arch Pathol Lab Med 2000;124:1135-40.
15. Rao LV, Jakubiak F, Sidwell JS, Winkelman JW, Snyder ML. Accuracy evaluation of a new glucometer with automated hematocrit measurement and correction. Clin Chim Acta 2005;356:178-83.
16. Mann EA, Salinas J, Pidcoke HF, Wolf SE, Holcomb JB, Wade CE. Error rates resulting from anemia can be corrected in multiple commonly used point-of-care glucometers. J Trauma 2008;64:15-21.
17. Pidcoke HF, Wade CE, Mann EA, Salinas J, Cohee BM, Holcomb JB, Wolf SE. Anemia causes hypoglycemia in intensive care unit patients due to error in single-channel glucometers: methods of reducing patient risk. Crit Care Med 2010;38:471-6.
18. Hoedemaekers CWE, Gunnewiek JM, Prinsen MA, Willems JL, Van der Hoeven JG. Accuracy of bedside glucose measurement from three glucometers in critically ill patients. Crit Care Med 2008;36:3062-6.
Procjena pogreške zbog hematokrita prilikom mjerenja koncentracije glukoze u krvi primjenom šest različitih prijenosnih glukometara
Uvod: Mjerenje i praćenje koncentracije glukoze u krvi kod hospitaliziranih bolesnika prijenosnim glukometrom (engl. portable glucose meters, PGMs) u širokoj je primjeni te predstavlja ključan dio liječenja šećerne bolesti, usprkos prisustvu nekoliko vrsta interferencija koje mogu uzrokovati negativnu pogrešku prilikom mjerenja. Svrha ovog istraživanja je procijeniti utjecaj hematokrita na analitički rad različitih PGM u usporedbi s referentnom laboratorijskom metodom.
Materijali i metode: Utjecaj različitih vrijednosti hematokrita (~0,20, ~0,45 i ~0,63 L/L) procijenjen je u tri uzorka pune krvi s različitom koncentracijom glukoze (~1,1, ~13,3 i ~25,0 mmol/L) primjenom šest različitih komercijalnih PGM. Isti su uzorci ispitivani referentnom laboratorijskom metodom (heksokinazom). Postotak razlike u odnosu na laboratorijsku metodu (%Diff) izračunat je na sljedeći način: %Diff = prosječna vrijednost dobivena PGM – vrijednost dobivena laboratorijskom metodom x 100 / vrijednost dobivena laboratorijskom metodom.
Rezultati: %Diff između šest različitih PGM bila je širokog raspona i iznosila između 56,5% do -34,8% kod uzoraka s niskom koncentracijom glukoze (~1,1 mmol/L), između 40% i -32% kod uzoraka s visokom koncentracijom glukoze (~13,3 mmol/L) i između -50% i 15% kod uzoraka s vrlo visokom koncentracijom glukoze (~5,0 mmol/L). Također je značajan podatak da je vrlo visoka vrijednost hematokrita (0,63 L/L) stvorila značajno negativnu pogrešku kod određivanja koncentracije glukoze u krvi (-35%) laboratorijskom metodom u usporedbi s referentnim uzorkom (hematokrit 0,45 L/L).
Zaključak: Rezultati ovog analitičkog istraživanja jasno potvrđuju da hematokrit stvara jaku i gotovo nepredvidivu pogrešku u primjeni PGM što uglavnom ovisi o različitom tipu uređaja. Zdravstveno osoblje i bolesnici moraju biti svjesni tog ograničenja pogotovo kod pojave ekstremne razine hematokrita kada se preporuča određivanje koncentracije glukoze u plazmi referentnim laboratorijskim metodama.
Ključne riječi: prijenosni glukometar; pogreška; hematokrit; analitički rad
Original scientific paper:
Giuseppe Lippi1*, Paola Avanzini1, Fernan da Pavesi1, Mirco Bardi1, Luigi Ippolito1, Rosalia Aloe1, Emmanuel J Favaloro2. Studies on in vitro hemolysis and utility of corrective formulas for reporting results on hemolyzed specimens. Biochemia Medica 2011;21(3):297-305. http://dx.doi.org/10.11613/BM.2011.040
1U.O. Diagnostica Ematochimica, Dipartimento di Patologia e Medicina di Laboratorio, Azienda Ospedaliero-Universitaria di Parma, Parma, Italy
2Department of Haematology, Institute of Clinical Pathology and Medical Research (ICPMR), Westmead Hospital, NSW, Australia
*Corresponding author: glippi [at] ao [dot] pr [dot] it, ulippi [at] tin [dot] it
Introduction: Spuriously hemolyzed specimens are the most common preanalytical problems in clinical laboratories. Corrective formulas have been proposed to allow the laboratory to release test results on these specimens. This study aimed to assess the influence of spurious hemolysis and reliability of corrective formulas.
Materials and methods:Blood collected into lithium heparin vacuum tubes was divided in aliquots and subjected to mechanical injury by aspiration with an insulin syringe equipped with a thin needle (30 gauge). Each aliquot (numbered from “#0” to “#5”) was subjected to a growing number of passages through the needle, from 0 to 5 times. After hematological testing, plasma was separated by centrifugation and assayed for lactate dehydrogenase (LD), aspartate aminotransferase (AST), potassium and hemolysis index (HI).
Results:Cell-free hemoglobin concentration gradually increased from aliquot #0 (HI: 0) to #5 (HI: 76±22, cell-free hemoglobin č 37.0 g/L). A highly significant inverse correlation was observed between HI and red blood cell count (RBC), hematocrit, mean corpuscular volume (MCV), LD, AST, potassium, whereas the correlation was negative with mean corpuscular hemoglobin (MCH). No correlation was found with hemoglobin, platelet count and glucose. A trend towards decrease was also observed for white blood cells count. The ANCOVA comparison of analyte-specific regression lines from the five subjects studied revealed significant differences for all parameters except potassium. In all circumstances the sy,x of these equations however exceeded the allowable clinical bias.
Conclusions:Mechanical injury of blood, as it might arise from preanalytical problems, occurs dishomogeneously, so that corrective formulas are unreliable and likely misleading.
Key words:hemolysis; hemolyzed specimens; interference; preanalytical variability
Laboratory diagnostics is a complex enterprise, with the ultimate aim to provide reliable and appropriate information to the key stakeholders, i.e., patients and clinicians. Although it is commonly assumed that laboratory testing is trustworthy and safe, a variety of risks might arise throughout the total testing process. Some of these problems can adversely affect the quality of results and furthermore waste precious economic healthcare resources, as well as jeopardize patient safety. The testing process is traditionally divided in three major stages, namely the preanalytical, analytical and postanalytical phases. Due to remarkable technological advances and strict internal and external control of the process, a major degree of safety has been achieved in the analytical phase, as granted by high test accuracy and precision. Accordingly, the major vulnerabilities seem to arise now from the pre- and postanalytical phases (1). In particular, considerable data has been gathered to support the notion that the manually intensive preanalytical activities are responsible for the vast majority of errors throughout the total testing process, and upwards of 70% of all errors (1-4). These activities broadly entail the collection, handling, transport, preparation and storage of the specimens that are subsequently utilized for testing.
Several individual studies and critical reviews of the literature have now clearly established that spuriously hemolyzed specimens represent the most common preanalytical problem encountered in the daily laboratory practice. Their frequency is comprised between 2 and 3% of all of routine samples, which is nearly five-time higher than the second cause (i.e., clotted samples). More importantly, only minorities of hemolyzed specimens reflect an in vivo cause of hemolysis (i.e., hemolytic anemia, occurring in č3% of all hemolyzed samples) (5). Thus, most of the hemolyzed samples referred to the laboratory for testing reflect the breakdown of erythrocytes and other blood cells during the collection, handling and/or transportation of the specimens. The multitude of factors that might generate spurious hemolysis typically begins during venipuncture and then continues downstream throughout this process, up to the time of analysis. These factors can be classified as follows (6,7):
- Factors dependent upon patient’s conditions (e.g., fragile veins or unsuitable venous access);
- Factors dependent upon the ability of the phlebotomist;
- Factors dependent upon the venipuncture per se (e.g., traumatic blood draw, unsatisfactory attempts, vein missing, prolonged placement of the tourniquet);
- Factors associated with the devices used for collecting the sample (e.g., syringes, cannulas, butterfly devices, small gauge needles);
- Factors dependent upon the conditions for transport (e.g., prolonged transportation, unsuitable environmental conditions - excess heat or cold – as well as contact of tubes with frozen packs);
- Factors dependent upon processing of the specimen (e.g., force, time and temperature of centrifugation, or generation of poor barrier integrity between blood cells and serum or plasma);
- Factors dependent upon the storage of specimens (e.g., refrigerated as whole blood with poor barrier between cells and sample, freezing-thawing of samples, storage in cyclic defrost freezers).
The major problem encountered with hemolyzed specimens is represented by the varying degree of interference with some laboratory assays, which is typically dependent upon biological factors (e.g., leakage of intracellular components into serum or plasma, release of cell derived tissue factor and phospholipids), chemical and/or spectrophotometric interference of cell-free hemoglobin in certain assays (6,7).
It has been previously highlighted that whenever a laboratory test is unreliable due to the presence of an “interferent” such as cell-free hemoglobin, the results should be suppressed and blinded to the clinicians (6-8). Nevertheless, this procedure is challenging, since it might cause relational problems, especially with the emergency department where the burden of hemolyzed specimens is comparatively high (9), or with emergency physicians who need urgent results for rapid patient triage (10). The advantages of routine determination of plasma or serum cell-free hemoglobin concentration have been described (11). Corrective formulas have also been proposed, so that adjustment of the data obtained on hemolyzed specimens would permit to release test results with an appropriate accompanying post-analytical comment (e.g., “results obtained on a hemolyzed specimen, suggest repeat testing for confirmation”) (12-14), or the hemolysis index (HI) (15-18). Some caveats have however been highlighted in this policy, including the imprecise estimation of the analytes following the use of these corrective equations, due primarily to the large intra-individual variability. Therefore, to further assess whether correction of hemolysis-sensitive laboratory tests for the degree of cell-free hemoglobin might be suitable, we performed a series of studies on mechanically-induced hemolyzed samples and potential corrective formulas for reporting results on these specimens. This is of notable importance since most clinical chemistry instruments are now equipped with serum indices, including the HI, which are much more reliable than classically applied visual inspection (19,20), and would permit an instantaneously applied correction of test results by incorporating a HI-based equation within specific instrumentation, middleware software or within the laboratory information system (LIS) (21).
Materials and methods
Samples and methods
Blood was collected by a single experienced phlebotomist early in the morning into 6.0 mL siliconized vacuum tubes without gel separator and containing 18 U/L lithium heparin (Vacuette, Greiner Bio-One GmbH, Frickenhausen, Germany), by using a 20 gauge, 0.80 x 19 mm straight needle (Greiner Bio-One GmbH). Two consecutive primary tubes were drawn from each of 5 healthy volunteers recruited among the laboratory staff (3 males and 2 females, mean age 42 years; range 38-46 years). The study was carried out according to the Declaration of Helsinki and under the terms of all relevant local legislation, and all subjects provided written informed consent. The blood from the two primary tubes was pooled. Spurious hemolysis was obtained by a variation of the original method of Dimeski (22), i.e., by aspirating the blood with a 0.5 mL insulin syringe equipped with a very thin needle (30 gauge, 0.3 x 8 mm). A first aliquot (“#0”) was separated from the rest of the blood and processed without further manipulation. A second aliquot (“#1”) was obtained by a single aspiration of the pooled blood. A third aliquot (“#2”) was then obtained by a single aspiration of the blood in aliquot #1, a fourth aliquot (#3) by a single aspiration of the blood in aliquot #2, a fifth aliquot (#4) by a single aspiration of the blood in aliquot #3, and a sixth aliquot (#5) by a single aspiration of the blood in aliquot #4. This method has been validated to reliably mirror a traumatic blood collection with production of a poor quality specimen (22), and is also expected to produce injury to platelets and white blood cells (WBC), other than to red blood cells (RBCs). The anticoagulated whole blood was immediately assessed on an Advia 2120 (Siemens Healthcare Diagnostics, Tarrytown NY, USA) for hematological testing. Lithium-heparin plasma from each aliquot was also subsequently obtained by centrifugation at 2000 x g for 15 min at room temperature, separation from the cell pellet and then tested for lactate dehydrogenase (LD; DGKC method), aspartate aminotransferase (AST; IFCC method with pyridoxal phosphate activation), potassium and HI on a Beckman Coulter DxC (Beckman Coulter Inc., Brea CA, USA), following manufacturer specifications and using proprietary reagents. The HI is assessed on Beckman Coulter DxC by direct spectrophotometry. Semiquantitative values are calculated on a linear scale from 0 (0 g/L of hemoglobin) to 10 (hemoglobin from 4.5 to 5.0 g/L). The highly significant correlation between HI and cell-free hemoglobin (measured with the reference cyanmethemoglobin assay) has been reported elsewhere (23). When outside the linearity of the methods, the aliquots were further diluted in saline to obtain a definitive value. Total imprecision (i.e., CV) of all the parameters tested has been reported to be < 2.5% for both the ADVIA 2120 (24), and Beckman Coulter DxC (25). The same instruments and reagent lots were used throughout the study and all measurements were performed within a single analytical session.
The comparison of analyte-specific regression lines obtained from each of the five subjects was done by ANCOVA. Correlation was calculated between the baseline MCV (mean corpuscular volume) value and the slope of the RBCs equation. The bias of results was also compared with the quality specifications derived from biologic variation for each of the parameter analyzed (26).
Statistical analysis was carried out by using Dsaastat for Excel version 1.1 (available at: http://www.unipg.it/~onofri/DSAASTAT/DSAASTAT.htm).
The main results of this investigation are reported in table 1. As expected, no significant amount of cell-free hemoglobin was found in the baseline aliquot #0 (i.e., HI of 0 in all samples, cell-free hemoglbin < 0.5 g/L). Conversely a progressive amount of cell-free hemoglobin was observed in the subsequent aliquots: aliquot #1, HI of 15±6 (cell free hemoglobin č7.0 g/L); aliquot #2, HI of 29±11 (cell free hemoglobin č14.0 g/L); aliquot #3, HI of 45±14 (cell free hemoglobin č22.0 g/L); aliquot #4, HI of 59±17 (cell free hemoglobin č29.0 g/L); and aliquot #5, HI of 76±22 (cell free hemoglobin č37.0 g/L). A highly significant inverse correlation was observed between HI and RBC count, hematocrit, MCV, LD, AST and potassium, whereas a highly significant negative correlation was observed with the Mean Corpuscular Hemoglobin (MCH). No significant correlation was instead observed between HI and hemoglobin, platelet count and glucose (Table 1). A marginally significant, negative correlation was observed between the HI and WBC count. In particular, although an evident trend was observed towards decreasing values of the WBC in parallel with the increase of the HI, the overall behavior was extremely heterogeneous among the study population, as shown in figure 1. Similarly, the platelet count also exhibited an apparent and remarkable increase in grossly hemolyzed specimens (i.e., HI > 100; cell-free hemoglobin of č50 g/L). The comparison by ANCOVA of analyte-specific regression lines obtained from each of the five subjects studied revealed that the resulting individual equations significantly differed (i.e., P < 0.001) among the five subjects for all parameters tested except potassium (P = 0.090) (Figure 2). A significant correlation was also observed between the baseline MCV value and the slope of the RBCs equation (r = -0.901; P = 0.037). As regards the clinical significance of these variations, in all circumstances the sy,x calculated from the individual linear regression analysis between the HI and the parameters, largely exceeded the quality specifications for desirable bias when expressed as percentage bias (Table 1).
Table 1. Correlation and linear regression analysis between the Hemolysis Index and results of hematological and clinical chemistry parameters in mechanically hemolyzed specimens collected from five healthy volunteers.
Figure 1. Behavior of white blood cell (WBC) (1A) and platelet (1B) count in mechanically hemolyzed specimens.
Figure 2. Linear regression analysis between hemolysis index (HI) and red blood cell count (2A), hemoglobin (2B), hematocrit (2C), MCV (2D), MCH (2E), LD (2F), AST (2G) and potassium (2H).
Hemolyzed specimens represent a crucial issue in laboratory diagnostics, both for the high frequency and for the important interference that cell-free hemoglobin, other intracellular components and cellular debris exert on a variety of clinical chemistry (28), coagulation (29), and immunochemistry assays (7), as well as on arterial blood gas analysis (30). Moreover, since hemolyzed specimens are often an important cause of relational, economical and organizational problems between laboratory professionals and physicians, especially those working in the emergency department (10), some corrective formulas have been proposed, since this might be beneficial for early triage and decision making, and further harmonizing professional relationships. At least hypothetically, when the lower bound of the predicted bias would provide a roughly acceptable value (i.e., within the reference range), the collection of a second sample might be unnecessary. In these specifically reported equations, the concentration of the analyte is multiplied by the slope obtained from a linear regression analysis between the bias observed at different cell-free hemoglobin concentrations, as assessed by either the cyanmethemoglobin assay (12,13), or the HI (15-17). Brescia et al. suggested an additional approach, where the analyte (i.e., potassium) released by lyzed RBCs is estimated from a formula including the MCHC along with cell-free hemoglobin (18). There are however several theoretical reasons proposed against the use of such formulas, which include (i) the large heterogeneity of the different formulas which prevents their interchangeability among different local conditions of instrumentation, assay and sample matrix (e.g., serum or plasma), (ii) the potentially broad interindividual variability of most intracellular constituents (e.g., potassium, LD and AST), and (iii) the misleading information about the biochemical profile in the presence of hemolytic anemia. While the last two aspects are virtually incontestable, the aim of this experiment was to verify whether also the third assumption was valid, thereby arguing definitively against the use of these equations.
Taken together, our results support the hypothesis that a dishomogenous interindividual behavior of clinical chemistry and hematological parameters exists following mechanical trauma of blood, so that the different tests could be essentially classified according to five patterns, as shown in figure 3, i.e., unaffected, homogenous interindividual variation and non clinically significant bias, heterogeneous interindividual variation and non clinically significant bias, homogenous interindividual variation and clinically significant bias, heterogeneous interindividual variation and clinically significant bias. The first and the second would reasonably reflect the ideal conditions. In the first case, the results might be safely released to the clinicians since the bias caused by blood cell lysis is virtually inappreciable. The second case might also be similarly handled, since the homogeneous interindividual variation of results and the lack of a clinically significant bias would allow the use of corrective formulas, calculated according to the concentration of cell-free hemoglobin in the sample to permit adjustment of test results accordingly. The three other conditions would instead lead to suppression of all test results and blinding of these to the clinicians since (i) one corrective formula would not be adequate for all cases in the presence of a high degree of interindividual variation of results, or (ii) the correction of results by any corrective equation would hypothetically produce an estimated value that exceeds clinical significance and would thereby be unreliable or misleading.
According to our results, hemoglobin and glucose fell within the first category, so that test results on hemolyzed specimens might be reliable regardless of the concentration of cell-free hemoglobin in the specimen (Figure 3). No parameter tested fell within the second category, i.e., displayed a homogenous interindividual variation and a non clinically significant bias that would permit the calculation of an analyte-specific, universal, corrective formula for each analyte to accurately estimate test results. Conversely, all remaining test parameters could be classified within the other three categories, where the use of an equation formula would be misleading due to the large interindividual variability of the values (RBC count, hematocrit, MCV and MCH), or the imprecise estimation since the percentage bias of the sy,x exceeded the quality specifications for desirable bias derived from biologic variation (i.e., potassium), or both aspects (platelet count, WBC count, LD and AST) (Figure 3) (26). In clinical practice, this implies that (i) the values of RBC count, hematocrit, MCV and MCH in hemolyzed specimens might be reliably predicted by a HI-based equation, which should however be individualized (i.e., separate equation for each patient); (ii) a common formula might be developed for predicting potassium values according to the HI, but the calculated value would not be sufficiently accurate to be clinically usable, and (iii) the values of platelet count, WBC count, LD and AST cannot be accurately predicted from the HI, nor can a common equation be identified. These results, at least for potassium, are in keeping with those of Shepherd et al., who correlated the bias between potassium in the hemolyzed and non-hemolyzed repeated samples with the HI (31). These workers also observed a significant linear relationship, mirrored by a 0.16 mmol/L potassium increase for each increment of the HI. Nevertheless, in agreement with our results, they observed a wide bias in potassium values calculated from the expected equation (i.e., ± 0.4 mmol/L vs. ± 0.5 mmol/L in our study), concluding that the magnitude of this variation was too excessive to recommend the use of the HI to predict potassium concentration in hemolyzed specimens.
Figure 3. Classification of hematological and clinical chemistry results in hemolyzed specimens.
WBC - white blood cells; RBC - red blood cells; Hb - hemoglobin; Ht – hematocrit; MCV - mean corpuscular volume; MCH - mean corpuscular hemoglobin; Plt – platelets; LD -lactate dehydrogenase; AST - aspartate aminotransferase.
Despite the limited number of subjects investigated, all the trends were however consistent. As such, an additional and important aspect that we observed in this study is that - although a similar pattern of RBCs lysis was observed among the study population – the slopes were statistically different one from the other, indicating that the different subjects were characterized by heterogeneous erythrocyte fragility. Interestingly, the slope of the RBCs equation was reliably predicted by the baseline MCV value (r = -0.901; P = 0.037), thereby suggesting that the differential degree of RBCs breakdown might at least partially depend upon the initial size of the erythrocytes, where the RBCs of subjects with smaller MCV tend to be more resistant to the mechanical lysis than those of subjects with a greater MCV. Unpredictable patterns were also observed for platelet count and WBC. While the results of the former parameter were dramatically unreliable, increasing progressively in parallel with the increase of the HI due to the well known interference from damaged blood cells and their cytoplasmic fragments in platelet enumeration and sizing (32), the behavior of WBC was roughly similar among the different individuals, displaying a significant inverse relationship with the HI and thereby confirming that WBCs other than RBCs might be dramatically injured in hemolyzed specimens.
In conclusion, the results of our investigation attest that the mechanical injury of the blood, as it might occur during flawed or mishandled procedures for collecting and handling blood specimens, does not occur homogeneously nor sufficiently predictably among different subjects, so that the use of corrective formulas to adjust and release test results on these samples is unreliable and likely to be even misleading.
Potential conflict of interest
1. Lippi G, Simundic AM. Total quality in laboratory diagnostics. It’s time to think outside the box. Biochem Med 2010;20:5-8.
2. Lippi G, Guidi GC, Mattiuzzi C, Plebani M. Preanalytical variability: the dark side of the moon in laboratory testing. Clin Chem Lab Med 2006;44:358-65.
3.Favaloro EJ, Lippi G, Adcock DM. Preanalytical and postanalytical variables: the leading causes of diagnostic error in hemostasis? Semin Thromb Hemost 2008;34612-34.
4.Lippi G, Chance JJ, Church S, Dazzi P, Fontana R, Giavarina D, et al. Preanalytical quality improvement: from dream to reality. Clin Chem Lab Med 2011;49:1113-26.
5.Carraro P, Servidio G, Plebani M. Hemolyzed specimens: a reason for rejection or a clinical challenge? Clin Chem 2000;46:306-7.
6. Lippi G, Blanckaert N, Bonini P, Green S, Kitchen S, Palicka V, et al. Haemolysis: an overview of the leading cause of unsuitable specimens in clinical laboratories. Clin Chem Lab Med 2008;46:764-72.
7. Simundic AM, Topic E, Nikolac N, Lippi G. Hemolysis detection and management of hemolysed specimens. Biochem Med 2010;20:154-9.
8. Dodig S. Interferences in quantitative immunochemical methods. Biochem Med 2009;19:50-62.
9. Lippi G, Salvagno GL, Favaloro EJ, Guidi GC. Survey on the prevalence of hemolytic specimens in an academic hospital according to collection facility: opportunities for quality improvement. Clin Chem Lab Med 2009;47:616-8.
10. Lippi G, Plebani M, Di Somma S, Cervellin G. Hemolyzed specimens: a major challenge for emergency departments and clinical laboratories. Crit Rev Clin Lab Sci 2011;48:143-53.
11. Koseoglu M, Hur A, Atay A, Cuhadar S. Effects of hemolysis interference on routine biochemistry parameters. Biochem Med 2011;21:79-85.
12. Owens H, Siparsky G, Bajaj L, Hampers LC. Correction of factitious hyperkalemia in hemolyzed specimens. Am J Emerg Med 2005;23:872-5.
13. Hawkins R. Variability in potassium/hemoglobin ratios for hemolysis correction. Clin Chem 2002;48:796.
14. Favaloro EJ, Lippi G. Laboratory reporting of haemostasis assays: The final post-analytical opportunity to reduce errors of clinical diagnosis in hemostasis? Clin Chem Lab Med 2010;48: 309-321.
15. Vermeer HJ, Steen G, Naus AJ, Goevaerts B, Agricola PT, Schoenmakers CH. Correction of patient results for Beckman Coulter LX-20 assays affected by interference due to hemoglobin, bilirubin or lipids: a practical approach. Clin Chem Lab Med 2007;45:114-9.
16. Jeffery J, Sharma A, Ayling RM. Detection of haemolysis and reporting of potassium results in samples from neonates. Ann Clin Biochem 2009;46:222-5.
17. Dimeski G, Clague AE, Hickman PE. Correction and reporting of potassium results in haemolysed samples. Ann Clin Biochem 2005;42:119-23.
18. Brescia V, Tampoia M, Mileti A. Evaluation of Factitious Hyperkalemia in Hemolytic Samples: Impact of the Mean Corpuscular Hemoglobin Concentration. LabMedicine 2009;40:224-6.
19. Simundic AM, Nikolac N, Ivankovic V, Ferenec-Ruzic D, Magdic B, Kvaternik M, Topic E. Comparison of visual vs. automated detection of lipemic, icteric and hemolyzed specimens: can we rely on a human eye? Clin Chem Lab Med 2009;47:1361-5.
20. Plebani M, Lippi G. Hemolysis index: quality indicator or criterion for sample rejection? Clin Chem Lab Med 2009;47:899-902.
21. Lippi G, Bonelli P, Rossi R, Bardi M, Aloe R, Caleffi A, Bonilauri E. Development of a preanalytical errors recording software. Biochem Med 2010;20:90-5.
22. Dimeski G. Effects of hemolysis on the Roche ammonia method for Hitachi analyzers. Clin Chem 2004;50:976-7.
23. Lippi G, Luca Salvagno G, Blanckaert N, Giavarina D, Green S, Kitchen S, et al. Multicenter evaluation of the hemolysis index in automated clinical chemistry systems. Clin Chem Lab Med 2009;47:934-9.
24. Harris N, Jou JM, Devoto G, Lotz J, Pappas J, Wranovics D, et al. Performance evaluation of the ADVIA 2120 hematology analyzer: an international multicenter clinical trial. Lab Hematol 2005;11:62-70.
25. Servonnet A, Théfenne H, Boukhira A, Vest P, Renard C. Evaluation of the Unicel DXC 600 (Beckman Coulter) and comparison with Integra 800 (Roche Diagnostics). Ann Biol Clin (Paris) 2007;65:555-62.
26. Ricos C, Alvarez V, Cava F, Garcia-Lario JV, Hernandez A, Jimenez CV, et al. Current databases on biologic variation: pros, cons and progress. Scand J Clin Lab Invest 1999;59:491-500.
27. Zar JH. Biostatistical Analysis, 2nd Edition. Prentice-Hall, Englewood Cliffs, USA (1984), pp. 176–179.
28. Lippi G, Salvagno GL, Montagnana M, Brocco G, Guidi GC. Influence of hemolysis on routine clinical chemistry testing. Clin Chem Lab Med 2006;44:311-6.
29. Lippi G, Montagnana M, Salvagno GL, Guidi GC. Interference of blood cell lysis on routine coagulation testing. Arch Pathol Lab Med 2006;130:181-4.
30. Lippi G, Ippolito L, Fontana R. Prevalence of hemolytic specimens referred for arterial blood gas analysis. Clin Chem Lab Med 2011;49:931-2.
31. Shepherd J, Warner MH, Poon P, Kilpatrick ES. Use of haemolysis index to estimate potassium concentration in in-vitro haemolysed serum samples. Clin Chem Lab Med 2006;44:877-9.
32. Zandecki M, Genevieve F, Gerard J, Godon A. Spurious counts and spurious results on haematology analysers: a review. Part I: platelets. Int J Lab Hematol 2007;29:4-20.
Istraživanja o in vitro hemolizi i primjeni korektivnih formula u izvještavanju o rezultatima iz hemolitičnih uzoraka
Uvod: Lagano hemolitični uzorci predstavljaju najčešći problem u prijeanalitičkoj fazi u radu kliničkih laboratorija. Predložene su korektivne formule koje omogućuju laboratoriju izdavanje rezultata iz takvih uzoraka. Cilj ovog istraživanja je procjena utjecaja hemolize in vitro na pouzdanost korektivnih formula.
Materijali i metode: Krv sakupljena u vakuum epruvete s litij-heparinom alikvotirana je te su u njoj mehanički oštećene stanice aspiracijom inzulinskom injekcijom s tankom iglom (30 gauge). Alikvoti su označeni brojevima od “#0” do “#5”. Ovisno o oznaci uzorci krvi su nakon vađenja (uzorak #0) provučeni su kroz iglu dodatnih 1 (uzorak “#1”) do 5 (uzorak “#5”) puta. Nakon hematološkog ispitivanja plazma je odvojena centrifugiranjem i u njoj je određena aktivnost laktat-dehidrogenaze (LD), aspartat-aminotransferaze (AST), koncentracija kalija i indeks hemolize (HI).
Rezultati: Koncentracija slobodnog hemoglobina postepeno je rasla od alikvota #0 (HI = 0) do #5 (HI = 76±22, slobodni hemoglobin č37,0 g/L). Primijećena je statistički značajna inverzna korelacija između HI i broja eritrocita, hematokrita, srednjeg volumena eritrocita (MCV), aktivnosti LDH, AST i koncentracije kalija, dok je korelacija s prosječnom količinom hemoglobina u eritrocitu (MCH) bila negativna, a nije bilo korelacije s koncentracijom hemoglobina i glukoze te brojem trombocita. Primijećen je trend prema nižoj koncentraciji leukocita. ANCOVA usporedbom regresijskih pravaca specifičnih za svaki analit dobivene su statistički značajne razlike za sve parametre osim kalija. Međutim, u svim oblicima je sy,x bio iznad dopuštene kliničke pogreške.
Zaključak: Mehaničko oštećivanje stanica u krvi koje se može dogoditi u prijeanalitičkoj fazi, nije homogeno, što čini korektivne formule nepouzdanima te je vrlo vjerojatno da će navesti na pogrešan zaključak.
Ključne riječi: hemoliza, hemolitični uzorci, interferencija, prijeanalitička varijabilnost