Clin Chem Lab Med 2014; 52(11): 1557–1568
Pilar Fernandez, María Antonia Llopis, Carmen Perich*, Maria Jesús Alsina, Virtudes Alvarez, Carmen Biosca, Gloria Busquets, Maria Vicenta Domenech, Rubén Gómez, Isabel Llovet, Joana Minchinela, Rosa Pastor, Rosa Ruiz, Ester Tarrés, Mercè Ibarz, Margarita Simón and Mercè Montesinos
Harmonization in hemolysis detection and prevention. A working group of the Catalonian Health Institute (ICS) experience Abstract Background: Hemolysis is the main cause of non-quality samples in clinical laboratories, producing the highest percentage of rejections in the external assurance programs of preanalytical quality. The objective was to: 1) study the agreement between the detection methods and quantification of hemolysis; 2) establish comparable hemolysis interference limits for a series of tests and analytical methods; and 3) study the preanalytical variables which most influence hemolysis production. Methods: Different hemoglobin concentration standards were prepared using the reference method. Agreement was studied between automated methods [hemolytic indexes (HI)] and reference method, as well as the interference according to hemolysis degree in various biochemical
*Corresponding author: Carmen Perich, Laboratori Clínic Bon Pastor, Mollerusa s/n, Barcelona 08030, Spain, E-mail: [email protected] Pilar Fernandez: Institut Català de la Salut (ICS), Laboratoris Clínics Vall d’Hebron, Barcelona, Spain María Antonia Llopis, Maria Jesús Alsina and Joana Minchinela: ICS, Laboratori Clínic Barcelonès Nord i Vallès Oriental, Barcelona, Spain Virtudes Alvarez and Rosa Ruiz: ICS, Laboratori Clínic l’Hospitalet, l’Hospitalet, Spain Carmen Biosca: Hospital Germans Trias i Pujol, Badalona, Spain Gloria Busquets and Mercè Montesinos: ICS, Laboratori Clínic Hospital Dr Josep Trueta, Girona, Spain Maria Vicenta Domenech: ICS, Laboratori Clínic Manso, Barcelona, Spain Rubén Gómez: Laboratorio Clínico Hospital La Paz, Madrid, Spain Isabel Llovet: ICS, Laboratori Clínic Hospital Verge de la Cinta, Tortosa, Spain Rosa Pastor: ICS, Laboratori Clínic Hospital Joan XXIII, Tarragona, Spain Ester Tarrés: ICS, Laboratori Clínic Bages, Berga, Spain Mercè Ibarz: ICS, Laboratori Clínic Hospital Arnau de Vilanova, Lleida, Spain Margarita Simón: Consorci Laboratori Intercomarcal, Vilafranca del Penedès, Spain
tests was measured. Preanalytical variables which could influence hemolysis production were studied: type of extraction, type of tubes, transport time, temperature and centrifugation conditions. Results: Good agreement was obtained between hemoglobin concentrations measured using the reference method and HI, for the most of studied analyzers, particularly those giving quantitative HI. The hemolysis interference cut-off points obtained for the majority of tests studied (except LDH, K) are dependent on the method/ analyzer utilized. Furthermore, discrepancies have been observed between interference limits recommended by the manufacturer. The preanalytical variables which produce a lower percentage of hemolysis rejections were: centrifugation at the extraction site, the use of lower volume tubes and a transport time under 15 min at room temperature. Conclusions: The setting of interference limits (cut-off) for each used test/method, and the study of preanalytical variability will assist to the results harmonization for this quality indicator. Keywords: hemolysis index; preanalytical errors; quality specifications; quality indicators. DOI 10.1515/cclm-2013-0935 Received October 31, 2013; accepted April 25, 2014; previously published online June 4, 2014
Introduction Hemolysis is the destruction process of red blood cells and entails the release of hemoglobin and other intra-erythrocytric components into plasma, altering its composition. Their presence could lead to an incorrect interpretation of analytical results with consequent repercussions for patient safety. Hemolysis constitutes one of the most
Brought to you by | SUNY Buffalo Libraries Authenticated Download Date | 10/13/14 3:21 PM
1558 Fernandez et al.: Harmonization between laboratories in hemolysis detection and interference limits frequent reasons for sample rejection in clinical chemistry laboratories, with a frequency range from 18.1% to 29.3%, as it is shown in the external preanalytic quality program reports [1, 2]. Numerous factors which could influence the production of hemolysis have been identified [3–6]. Their identification and correction could decrease this potential error source. They are classified in the following types: –– During the phlebotomy: extraction system (vacuum tubes instead of syringes), type of vascular access device (catheters instead of needles), gauge needle, arm position, syringe puncture and vacuum tube transfer, vigorous shaking or non-shaking of the sample after collection, skill of phlebotomist, characteristics of the patients related to blood vessels, etc. –– During transport: position of the tubes (horizontal or vertical), type of transport (pneumatic tube), the transport time and temperature. –– During the sample preparation: type of centrifugation, the time between blood collection and centrifugation, temperature, speed and centrifugation time, failures in the separation barrier, centrifugation of partially coagulated samples, re-centrifugation, etc. Detection of hemolysis in the laboratory is carried out by visual inspection or by automated analyzers that estimate the concentration of hemoglobin in the sample by measurement of absorbance at different wavelengths. Currently, most laboratories use automated systems. Visual detection is a subjective appraisal dependent on technician experience and hence the detection of hemolysis when hemoglobin concentration is below 0.2 g/L is difficult, due to the presence of other pigments, such as bilirubin, carotene, etc. A visually detectable minimum concentration of 0.3 g/L is being considered. The inclusion of automated hemolysis detection systems in clinical biochemical analyzers implies the removal of the subjective component, making easier the results harmonization among different laboratories. Since 2004, the group of clinical laboratories of the Institut Català de la Salut, (Catalonian Health Institute) have been evaluating the results obtained from different quality indicators, with the aim of obtaining firm quality specifications for all the processes, including the preanalytical phase processes [7–10]. To define the quality specifications of these indicators, results were chosen among the laboratory comparison results (state of the art) [11]. Regarding the indicator of hemolyzed samples, the specification obtained in this working group was 0.6%, calculated as the number of samples which invalidated
results for one or more requested tests, with regards to the number of demands. However, there was a great interlaboratory variability, ranging from 0.01% to 1.9% [10], mainly due to the following facts: the lack of standardization in hemolysis detection methods, the existence of non-homogeneous interference limits and the different characteristic of the samples collection centers. The presence of these variables, which correspond to different processes carried out in- and outside the laboratory, made it difficult to compare results for this indicator among the laboratories in the group, with the objective of obtaining quality specifications. The aim of this research was to study some factors that could have been the cause of variability in the results for the indicators. In order to develop this project, the working group set up a multi-centric study, divided into three phases: 1st phase: To study the agreement between the detection and quantification methods of hemolysis used by participant laboratories, comparing their results with those obtained by using the reference method for measuring the concentration of hemoglobin in plasma. 2nd phase: To evaluate the effect of the hemolysis degree in the results of several tests, which were measured using different analytical methods in order to establish comparable interference limits. 3rd phase: To study the influence of preanalytical variables in the hemolysis mechanism.
Materials and methods The working group was made up of 12 public laboratories: seven hospital laboratories (6 of them also with primary care) and five laboratories only with primary care belonging to the Institut Català de la Salut (ICS; Catalonian Health Institute). The mentioned laboratories are located in Catalonia, a region on the northeastern of Spain, within a radius of approximately 150 km. For the first phase of the study, quantification of the free hemoglobin concentration in serum samples was carried out, according to the adapted reference method of cyanmethemoglobin that uses potassium ferricyanide and potassium cyanide to transform hemoglobin into cyanmethemoglobin, which showed a maximum absorption at 540 nm of wavelength [12]. This method had to be adapted in order to quantify hemoglobin concentrations within the interest interval (from 0 to 10 g/L), preparing two calibration curves, one from 2 to 10 g/L and the other one from 0 to 2 g/L of hemoglobin. A spectrophotometer UV-VIS was used (Shimadzu, Kyopto, Japan) with wavelength selection at 540 nm, reactive dilution Drabkin (Sigma-Aldrich, St. Louis, MO, USA), Solution Brj35 (Sigma®) as non-ionic detergent and human lyophilized hemoglobin (Sigma®). Samples were prepared from a serum pool without apparent hemolysis. In addition a pool of whole blood samples, collected in a tube with lithium heparin, was done. Three freezing and de-freezing
Brought to you by | SUNY Buffalo Libraries Authenticated Download Date | 10/13/14 3:21 PM
Fernandez et al.: Harmonization between laboratories in hemolysis detection and interference limits 1559 cycles (at –70 °C) for the red blood cells lysis and later centrifugation at 1500 g for 10 min were done in order to obtain an hemolyzed pellet. Serial dilutions were carried out by adding this pellet to serum pool to obtain samples within the following theoretical concentration of hemoglobin: 0; 0.16; 0.30; 0.57; 0.97; 2.43; 4.85 and 6.93 g/L (target values). The hemoglobin measurement and sample dilutions were carried out in one reference laboratory of the group. The samples were collected in the reference laboratory the day before to carry out the tests and then they were taken to every laboratory. The reference laboratory location allowed that transport time to other laboratories was not longer than 1 h. For the automated detection, HI were evaluated, in triplicate, by the following analyzers: four Cobas 711 analyzers (Hoffmann-La Roche Ltd, Basel, Switzerland), three AU 5400 (Beckman Coulter Inc, Fullerton, CA, USA), one Advia 2400 analyzer (Siemens Healthcare Diagnostics Inc, Deerfield, IL, USA), one DXC 800 (Beckman), one Synchron LXi725 (Beckman), one Cobas 6000 (Roche), one Modular DP (Roche), one Vista (Siemens). Different ways to express HI detection for these analyzers were summarized in Table 1. To evaluate the effect of the hemolysis degree in the results for different analytes (second phase), all the samples were analyzed in triplicate under same operating conditions for the following tests: alanine aminotransferase: (ALT), aspartate aminotransferase (AST), creatine kinase (CK), cholesterol (COL), phosphorus (P), alkaline phosphatase (ALP), iron (FE), glucose (GLU), γ-glutamyl-transferase (GGT), potassium (K), lactate dehydrogenase (LD), total protein (TP), triglycerides (TG), using all the mentioned analyzers. The analytical methods used to measure the different analytes, are shown in Table 2. The average of the three carried out determinations, was calculated for each hemoglobin standard and for each test. For each test, the obtained value in the standard of hemoglobin concentration 0 g/L was taken as baseline and the differences in percentage were calculated between the results obtained in each standard in comparison to the baseline. The cut-off point to define hemolysis interference for a specific test in each laboratory was established, when the percentage surpassed the bias derived from biological variation database [13, 14].
For each analyzer the interference limit was established as the median of the cut-off points obtained in each laboratory. To study the most influential variables in hemolysis production (third phase), the percentage of K result rejection in regard to those requested results, was calculated, considering rejection when the hemolysis exceeded the interference limit obtained in the second phase of the study for this test. For 1 month, in 324 primary care blood collection centers, where the samples transport (in refrigerators or at room temperature) was carried out by road to seven of the participant laboratories, the following data were collected as critical points: 1. Extraction system, (syringe or vacuum system) taking into account the provider and the sample volume for each tube; BD Vacutainer blood collection tubes (3.5 mL and 8.5 mL) (Becton Dickinson, NJ, USA) and Vacuette blood collection tubes (4 mL and 10 mL) (Vacuette, Greiner Bio-one, Kremsmuenster, Austria). 2. Centrifugation site (in the blood collection center or in the laboratory) and centrifugation conditions. 3. Transport time. 4. Temperature.
Statistical analysis Weighted κ coefficient (http://www.vassarstats.net/kappa.html) and intra-class correlation coefficient (ICC) were used for comparability of automated detection of hemolysis in regard to reference method. A weighted κ coefficient above 0.8 and an ICC above 0.75 were considered as a good agreement [15]. A forward multiple regression analysis has been carried out to assess the effect of such preanalytical characteristics upon the rejection percentage, so that the equation turns to be defined by those variables automatically included and their associated coefficients. The odds ratio was calculated – for each level – dividing the percentage of rejected specimens by the baseline percentage of rejected specimens (the baseline level for each variable was the level occurring most often). The transport time variable was divided into several intervals and the odds ratio was calculated with regards to a time interval 0–15 min.
Table 1 Different ways to express hemolytic indexes for studied analyzers.
Result
Beckman
Synchron Lxi725–DXC800
Qualitative 0 Not detected 1 Hb = (0–0.5) 2 Hb = (0.5–1) 3 Hb = (1–1.5) 4 Hb = (1.5–2) 5 Hb = (2–2.5) 6 Hb = (2.5–3) 7 Hb = (3–3.5) 8 Hb = (3.5–4) 9 Hb = (4–4.5) 10 Hb = (4.5–5)
Equivalence range hemoglobin, g/L
Siemens
AU 5400
Vista
Qualitative
Qualitative
0 (Hb ≤ 0.5) + (0.5 5 > 2– ≤ 3 > 0.5– ≤ 2 1.90 ± 0.01 1.83 ± 0.01 1.88 ± 0.02 1.80 ± 0.00 1.91 ± 0.02 1.74 ± 0.12 1.82 ± 0.02
> 5 > 5 > 5 > 5 > 5 > 3– ≤ 5 3.76 ± 0.01 3.57 ± 0.02 3.67 ± 0.01 3.52 ± 0.03 3.70 ± 0.04 3.37 ± 0.02 3.54 ± 0.02
> 5 > 5 ABNa ABNa ABNa > 3– ≤ 5 5.18 ± 0.02 4.99 ± 0.04 5.13 ± 0.03 4.91 ± 0.03 5.19 ± 0.04 4.73 ± 0.05 4.96 ± 0.02
a ABN, sample flagged as abnormal by the AU 5400 reagent; bHI qualitatives, In all cases the three observations were equivalent and expressed as an interval; cHI quantitatives, for the different analyzers expressed as mean ± SD. HI, hemolytic indexes.
Brought to you by | SUNY Buffalo Libraries Authenticated Download Date | 10/13/14 3:21 PM
1562 Fernandez et al.: Harmonization between laboratories in hemolysis detection and interference limits Table 4 κ of Cohen index and coefficients of intra-class correlation.
Beckman Synchron Lxi725 –DXC800
κ CI 95%
0.7895 0.6899–0.8891
AU 5400
0.8326 0.7278–0.9374
Siemens Vista
0.8209 0.6587–0.9831
Discussion There is a great variability in the percentage of hemolyzed samples in different laboratories, in a range between 0.05% and 3.3% according to different authors. The variation sources could be due to the different ways of obtaining and preparing samples, the different methods used by laboratories to detect and quantify hemolysis, the different criteria to set rejection interference limits for hemolyzed samples, as well as the different formulas used to calculate the indicator, which could be related to the number of blood samples, total samples, serum tubes or requests [3]. The final aim of our study was to evaluate preanalytical variables with most influential in the hemolysis production, after setting homogeneous interference limits for determined tests in all participant laboratories, as well as to establish the same rejection criteria and calculation formula. The reference method adapted for measurement of the hemoglobin concentration allowed us to quantify concentrations below 1 g/L, with an acceptable analytical performance for the study objective.
Siemens
CCI CI 95%
0.982 0.910–0.996
Advia
Roche Cobas-Modular
0.973 0.863–0.995
Visual hemolysis detection made difficult to carry out comparative studies among laboratories, due to the fact that this process was a low sensitive and subjective assessment which depends on the laboratory staff. In this sense, Glick et al. [17] and Simundic et al. [18] have reported disagreement among concentrations assigned visually and the concentration of real hemoglobin of a series of patterns and automated HI detection, respectively. There were a few studies comparing hemoglobin concentrations measured by the reference method to the HI obtained in different analyzers. Our results showed a high agreement for the most studied analyzers, similar to the results obtained by Lippi et al. [19]. The highest agreement regarding to reference method was obtained in those analyzers which give quantitative information about HI, although the results were undervalued, the homogeneity was allowed for all the analytical systems to set up a common indicator. According to our results, for some tests such as LDH, K and P (the majority released in the hemolysis process) homogeneous interference limits were obtained
Table 5 Proposed hemolysis interference limit (median per group of analyzers), (A) versus limit recommended by the manufacturer (B). Test
Hemolysis interference limit, g/L Advia 2400 (Siemens) A
ALT ASTa CKa COL P FAL FEa GLUa GGTa K LDa PTa TG a
2.4 0.6 2.4 > 6.9 2.4 2.4 4.8 2.4 6.9 0.6 0.16 6.9 6.9
B 5 5 5 7.5 5.25 5 na 5.25 5 na na 5 5.25
Synchron LXi725–DXC 800 (Beckman)
AU 5400 (Beckman)
A
B
A
1.7 0.3 1.7 > 6.9 3.6 4.8 0.6 > 6.9 3.6 0.6 0.16 3.6 3.6
0.5 0.5 0.5 5 1.5 5 0.5 5 2.5 0.5 0.5 5 5
2.4 0.6 4.8 2.4 2.4 2.4 6.9 > 6.9 6.9 0.6 0.16 2.4 6.9
B 5 na 1 5 3.5 4.5 1 na 5 na na 3 5
Cobas 711,6000 (Roche) A
Test with different analytical methods according to the analyzers. na, not available.
a
Brought to you by | SUNY Buffalo Libraries Authenticated Download Date | 10/13/14 3:21 PM
2.4 0.4 2.4 > 6.9 2.4 4.8 2.4 > 6.9 4.8 0.6 0.16 2.4 4.8
B 2 0.4 2 7 3 2 2 10 2 na 0.15 10 7
Vista (Siemens) A
B
4.8 0.2 4.8 4.8 2.4 4.8 2.4 0.9 > 6.9 1 0.16 0.9 > 6.9
> 10 0.25–0.5 5–10 > 10 5–10 > 10 0.25–0.5 > 10 5–10 0.25 0.1–0.25 > 10 > 10
Fernandez et al.: Harmonization between laboratories in hemolysis detection and interference limits 1563
AST. % Dif f erences between the results obtained vs baseline (22.4 U/L)
ALT. % Dif f erences between the results obtained vs baseline (18.1 U/L) 175.0
35.0 30.0
Beckman (Synchron Lxi 725/ DXC 800)
150.0
25.0
Beckman AU
125.0
20.0
Siemens (Vista)
15.0
Siemens (Advia 2400)
10.0
Roche (Cobas/Modular)
5.0
Optimal Spec (5.7%)
0.0 0.16
0.3
0.57
0.97
2.43
4.85
Beckman (Synchron Lxi 725/ DXC 800) Beckman AU
100.0
Siemens (Vista)
75.0
Siemens (Advia 2400)
50.0
Roche (Cobas/Modular)
25.0
Desirable Spec (5.4%)
0.0 0.16
6.93
0.3
0.57
0.97
2.43
4.85
6.93
Hemoglobin added, g/L
Hemoglobin added, g/L CK. % Dif f erences between the results obtained vs baseline (137.5 U/L)
Cholesterol. % Dif f erences between the results obtained vs baseline (209.4 mg/dL)
35.0
15.00
30.0
Beckman (Synchron Lxi 725/ DXC 800)
25.0
Beckman AU
20.0
Beckman (Synchron Lxi 725/ DXC 800)
10.00
Beckman AU
5.00
Siemens (Vista)
Siemens (Vista)
15.0
Siemens (Advia 2400)
10.0
Roche (Cobas/Modular)
5.0
0.00
Siemens (Advia 2400)
-5.00
Roche (Cobas/Modular) Desirable Spec (4%)
-10.00
Optimal Spec (5.8%)
0.0 0.16
0.3
0.57
0.97
2.43
4.85
-15.00
6.93
0.16
0.3
Hemoglobin added, g/L
0.57
0.97
2.43
4.85
6.93
Hemoglobin added, g/L
Alkaline Phosphatase. % Dif f erences between the results obtained vs baseline (69.1 U/L)
Phosphate. % Dif f erences between the results obtained vs baseline (4.5 mg/dL)
15.00
10.00 Beckman (Synchron Lxi 725/ DXC 800)
10.00
Beckman AU
5.00
Beckman (Synchron Lxi 725/ DXC 800)
0.00
Beckman AU
-5.00 Siemens (Vista)
5.00
Siemens (Vista)
-10.00 Siemens (Advia 2400)
0.00
-5.00
-15.00
Siemens (Advia 2400)
Roche (Cobas/Modular)
-20.00
Roche (Cobas/Modular)
Desirable Spec (3.2%)
-25.00
Desirable Spec (6.4%)
-30.00 0.16
0.3
0.57 0.97 2.43 4.85 Hemoglobin added, g/L
6.93
0.16
0.3
0.57 0.97 2.43 4.85 Hemoglobin added, g/L
6.93
(Figure 1 Continued)
Brought to you by | SUNY Buffalo Libraries Authenticated Download Date | 10/13/14 3:21 PM
1564 Fernandez et al.: Harmonization between laboratories in hemolysis detection and interference limits Glucose. % Dif f erences between the results obtained vs baseline (99.4 mg/dL )
Iron. % Dif f erences between the results obtained vs baseline (98.4 µg/dL ) 10.00
35.00 30.00
Beckman (Synchron Lxi 725/ DXC 800)
25.00
Beckman AU
20.00 15.00
Siemens (Vista)
10.00
Siemens (Advia 2400)
Beckman (Synchron Lxi 725/ DXC 800)
5.00
Beckman AU
0.00 Siemens (Vista)
-5.00
5.00
Siemens (Advia 2400)
-10.00
Roche (Cobas/Modular)
0.00
Optimal Spec (4.4%)
-5.00
Roche (Cobas/Modular)
-15.00
Desirable Spec (2.2%)
-20.00
-10.00 0.16
0.3
0.57
0.97
2.43
4.85
0.16
6.93
0.3
Hemoglobin added, g/L
0.57
0.97
2.43
4.85
6.93
Hemoglobin added, g/L
GGT. % Dif f erences between the results obtained vs baseline (28.0 U/L)
Potassium. % Dif f erences between the results obtained vs baseline (4.3 mmol/L) 30.00
15.00 10.00
Beckman (Synchron Lxi 725/ DXC 800)
5.00
Beckman AU
0.00
Beckman (Synchron Lxi 725/ DXC 800)
25.00
Beckman AU
20.00
Siemens (Vista)
Siemens (Vista)
15.00
-5.00 -10.00
Siemens (Advia 2400)
-15.00
Roche (Cobas/Modular)
-20.00
Optimal Spec (5.4%)
-25.00
0.16
0.3
0.57
0.97
2.43
4.85
Siemens (Advia 2400)
10.00
5.00
Roche (Cobas/Modular)
0.00
Minimal Spec (2.8%)
0.16
6.93
0.3
0.57
0.97
2.43
4.85
6.93
Hemoglobin added, g/L
Hemoglobin added, g/L
T. Protein. % Dif f erences between the results obtained vs baseline (7.5 g/L)
LD. % Dif f erences between the results obtained vs baseline (330 U/L) 15.00
300.00 Beckman (Synchron Lxi 725/ DXC 800)
250.00
Beckman (Synchron Lxi 725/ DXC 800)
10.00
Beckman AU
Beckman AU
200.00
Siemens (Vista)
5.00 Siemens (Vista)
150.00
100.00
Siemens (Advia 2400)
50.00
Roche (Cobas/Modular) Desirable Spec (4.3%)
0.00
0.16
0.3
0.57 0.97 2.43 4.85 Hemoglobin added, g/L
6.93
Siemens (Advia 2400)
0.00
Roche (Cobas/Modular)
-5.00
Minimal Spec (1.8%)
-10.00
0.16
0.3
0.57 0.97 2.43 4.85 Hemoglobin added, g/L
6.93
(Figure 1 Continued)
Brought to you by | SUNY Buffalo Libraries Authenticated Download Date | 10/13/14 3:21 PM
Fernandez et al.: Harmonization between laboratories in hemolysis detection and interference limits 1565
15.00
Triglyceride. % Differences between the results obtained vs baseline (133.7 mg/dL)
Beckman (Synchron Lxi 725/ DXC 800)
10.00 5.00
Beckman AU
0.00
Siemens (Vista)
-5.00
Siemens (Advia 2400)
-10.00
Roche (Cobas/Modular)
-15.00
Optimal Spec (5.3%)
-20.00 0.16
0.3
0.57 0.97 2.43 4.85 Hemoglobin added, g/L
6.93
Figure 1 Percentage difference for each standard in comparison to the baseline.
among the different analyzers. However, for the rest of the analyzed tests, the hemolysis interference cut-off point depended on the method and/or analyzer studied (Table 5). There were discordant results among analyzers for some parameters, e.g., Fe, with an interference limit between 0.6 and 6.9 g/L or GLU with a limit between 0.9 and > 6.9 g/L depending on the analyzer. When these results were compared with those published in different studies, discrepancies were also observed in interference limits for different tests [14, 20]. As has been mentioned previously in the Introduction, hemoglobin concentrations below 0.3 g/L are not easily detected by visual inspection; for this reason, hemolysis detection would be advisable in the analyzers for those analytes (LDH, AST and K), which showed an interference
limit below the minimum concentration visually detected [14]. In the same way, it is advisable that all analyzers detect the previous concentrations. According to our results, it has been observed that analyzers Beckman AU 5400, DXC800 and Synchron LXi725, which carried out a qualitative assessment of the HI, were not able to detect hemoglobin concentrations below 0.5 or 1 g/L and, therefore, significant interferences were not detected for some tests. As previously exposed, the implementation of HI in analyzers is advisable, because it also has been recently demonstrated by Lippi et al. [21] that systematical measurement of HI does not impair instrument efficiency. There is not universal consensus to set up the maximum allowable bias due to hemolysis [22]. The allowable error for interference based on the biological variation of the analyte [23] defines an interference clinically significant, when the result in the presence of the interferent differs more than 1.96 × (CVa2+CVw2)½ from the result without the interferent, where CVa is the analytical coefficient of variation and CVw the within-subject biological variation. The IFCC Working Group on Allowable Errors for Traceable Results [24] has stated that this bias should not exceed a definite part of total error uncertainty (total error budget) of a measurement. The considered criterion in this research as well as, Lippi et al. [14] to establish the thresholds for hemolysis interference for each measured analyte, was defined on the basis of comparison of desirable bias derived from biological variation [13]. This criterion has narrowed the analytical variation allowable only to the biological variation, and we believe that is a way to detect the interference for hemolysis with higher sensitive.
Table 6 Rejection and odds percentages calculated for the considered preanalytical variables. Preanalytical variables
Centers, n
Rejections, n
Serum tubes, n
Rejections, %
Odds (L.Inf-L.Sup CI 95%)
Transport time 0–15 min Transport time 16–60 min Transport time 61–120 min Transport time 120–150 min Transport at room temperature Transport at cooled temperature Centrifugation in laboratory Centrifugation in blood collection center Tubes vacuette Tubes BD-vacutainer Tubes volume 3.5 and 4 mL Tubes volume 8, 8.5 and 9 mL Vacuum extraction system Syringe extraction system
24 153 126 18 155 164 312 12 90 234 221 103 319 3
210 802 536 28 792 708 1552 21 610 963 1109 464 1543 22
11,154 32,608 19,832 982 39,019 20,658 62,972 1775 24,351 40,396 48,575 16,172 63,968 416
1.88 2.46 2.70 2.85 2.03 3.43 2.46 1.18 2.51 2.38 2.28 2.87 2.41 5.29
a
Baseline variable.
a
Brought to you by | SUNY Buffalo Libraries Authenticated Download Date | 10/13/14 3:21 PM
1.306 (1.124–1.518) 1.436 (1.289–1.599) 1.514 (1.041–2.202) a
1.688 (1.528–1.866) a
0.480 (0.313–0.736) 1.051 (0.950–1.162) a a
1.257(1.129–1.398) a
2.192 (1.455–3.302)
1566 Fernandez et al.: Harmonization between laboratories in hemolysis detection and interference limits In the in vitro samples used in our study, the influent variation was due to analytical inaccuracy in addition to hemolysis interference. The influence of biological variation was avoided by calculating the percentual difference among results of each standard with regards to basal value. Carrying out determinations in the same analytical batch kept the variability due to analytical bias as low as possible. Finally, carrying out serial determinations, the analytical imprecision was kept in a low level. For this reason the percentage of variation regarding baseline will be due to hemolysis interference. Discrepancies were also observed between the interference limits recommended by manufacturers and those obtained in our study. It is worth noting, the difference observed in cut-off used by different manufacturers. As was shown in Table 5 for phospate, the interference limit range among manufacturers was 1.5–10 g/L, however, the interference limit range for hemolysis in our results for different equipment ranged from 2.4 to 3.6 g/L. These discrepancies were probably due to differences in the criteria chosen to establish the cut-off point, which makes comparison difficult. In our case the cut-off was established when the difference (in percentage) between the result of the test on each standard and the baseline (HI: 0 g/L) exceeded the systematic error based on biological variation. In this sense manufacturers should make an effort to supply the complete data for the interference study (interferograms), being shown the type of study and the methodology utilized for the hemolysis quantification, as well as providing indexes which can discriminate free hemoglobin concentrations below 0.5 g/L. According to our results, it would be advisable for each laboratory to verify the information supplied by providers with regards hemolysis interference and to set up interference limits for those biochemical tests susceptible to be interfered, according to their own quality specifications. So far, no manufacturer has supplied calibrators and controls for serum indexes (hemolytic, icteric or lipemic), making difficult to control these methods. The inclusion of internal and external quality controls for these automated detection methods of serum index would mean an advance in standardization and results inter-comparison. According to our results, COL, GLU, GGT and TGs have an interference limit for hemolysis higher than 6.9 g/L, in some studied analyzers. At the moment of carrying out the study, a high volume of sample was requested (from Blood Bank) in order to supply aliquots of all the standards to the 12 participant laboratories of the study. Despite the recommended maximum concentration of hemoglobin to be tested to establish the
hemolysis interference is 10 g/L [25], due to the percentage of hemolyzed samples with a high hemolysis degree ( > 6.9 g/L) received in the laboratory is less than the total percentage of hemolyzed samples, a new concentration limit of hemoglobin in 6.9 g/L was suggested. As an example, the percentage of samples with a hemolysis degree higher than 6 g/L was calculated for 3 months in one of the participant laboratory groups and a value of 0.02% of a total 81,390 hemolyzed samples was obtained. The last phase of our study was approached once homogeneous interference limits had been set up for each test in every involved laboratory. In this phase, rejection percentages obtained for K analyte were calculated in order to determine the preanalytical variable with the most influence on the in vitro hemolysis production. Of all the variables which could influence hemolysis of samples, those considered most relevant and which could be collected without error by all the participant laboratories were selected (Table 6). In most cases centrifugation was carried out in the laboratory and in some further centers the centrifugation was carried out in the same collection center (centrifugation in origin), but in this case the samples were sent to the laboratory after centrifugation. The hemolysis rejections percentage has been increased significantly depending on transport time from 1.88%, if transport time was
Pilar Fernandez, María Antonia Llopis, Carmen Perich*, Maria Jesús Alsina, Virtudes Alvarez, Carmen Biosca, Gloria Busquets, Maria Vicenta Domenech, Rubén Gómez, Isabel Llovet, Joana Minchinela, Rosa Pastor, Rosa Ruiz, Ester Tarrés, Mercè Ibarz, Margarita Simón and Mercè Montesinos
Harmonization in hemolysis detection and prevention. A working group of the Catalonian Health Institute (ICS) experience Abstract Background: Hemolysis is the main cause of non-quality samples in clinical laboratories, producing the highest percentage of rejections in the external assurance programs of preanalytical quality. The objective was to: 1) study the agreement between the detection methods and quantification of hemolysis; 2) establish comparable hemolysis interference limits for a series of tests and analytical methods; and 3) study the preanalytical variables which most influence hemolysis production. Methods: Different hemoglobin concentration standards were prepared using the reference method. Agreement was studied between automated methods [hemolytic indexes (HI)] and reference method, as well as the interference according to hemolysis degree in various biochemical
*Corresponding author: Carmen Perich, Laboratori Clínic Bon Pastor, Mollerusa s/n, Barcelona 08030, Spain, E-mail: [email protected] Pilar Fernandez: Institut Català de la Salut (ICS), Laboratoris Clínics Vall d’Hebron, Barcelona, Spain María Antonia Llopis, Maria Jesús Alsina and Joana Minchinela: ICS, Laboratori Clínic Barcelonès Nord i Vallès Oriental, Barcelona, Spain Virtudes Alvarez and Rosa Ruiz: ICS, Laboratori Clínic l’Hospitalet, l’Hospitalet, Spain Carmen Biosca: Hospital Germans Trias i Pujol, Badalona, Spain Gloria Busquets and Mercè Montesinos: ICS, Laboratori Clínic Hospital Dr Josep Trueta, Girona, Spain Maria Vicenta Domenech: ICS, Laboratori Clínic Manso, Barcelona, Spain Rubén Gómez: Laboratorio Clínico Hospital La Paz, Madrid, Spain Isabel Llovet: ICS, Laboratori Clínic Hospital Verge de la Cinta, Tortosa, Spain Rosa Pastor: ICS, Laboratori Clínic Hospital Joan XXIII, Tarragona, Spain Ester Tarrés: ICS, Laboratori Clínic Bages, Berga, Spain Mercè Ibarz: ICS, Laboratori Clínic Hospital Arnau de Vilanova, Lleida, Spain Margarita Simón: Consorci Laboratori Intercomarcal, Vilafranca del Penedès, Spain
tests was measured. Preanalytical variables which could influence hemolysis production were studied: type of extraction, type of tubes, transport time, temperature and centrifugation conditions. Results: Good agreement was obtained between hemoglobin concentrations measured using the reference method and HI, for the most of studied analyzers, particularly those giving quantitative HI. The hemolysis interference cut-off points obtained for the majority of tests studied (except LDH, K) are dependent on the method/ analyzer utilized. Furthermore, discrepancies have been observed between interference limits recommended by the manufacturer. The preanalytical variables which produce a lower percentage of hemolysis rejections were: centrifugation at the extraction site, the use of lower volume tubes and a transport time under 15 min at room temperature. Conclusions: The setting of interference limits (cut-off) for each used test/method, and the study of preanalytical variability will assist to the results harmonization for this quality indicator. Keywords: hemolysis index; preanalytical errors; quality specifications; quality indicators. DOI 10.1515/cclm-2013-0935 Received October 31, 2013; accepted April 25, 2014; previously published online June 4, 2014
Introduction Hemolysis is the destruction process of red blood cells and entails the release of hemoglobin and other intra-erythrocytric components into plasma, altering its composition. Their presence could lead to an incorrect interpretation of analytical results with consequent repercussions for patient safety. Hemolysis constitutes one of the most
Brought to you by | SUNY Buffalo Libraries Authenticated Download Date | 10/13/14 3:21 PM
1558 Fernandez et al.: Harmonization between laboratories in hemolysis detection and interference limits frequent reasons for sample rejection in clinical chemistry laboratories, with a frequency range from 18.1% to 29.3%, as it is shown in the external preanalytic quality program reports [1, 2]. Numerous factors which could influence the production of hemolysis have been identified [3–6]. Their identification and correction could decrease this potential error source. They are classified in the following types: –– During the phlebotomy: extraction system (vacuum tubes instead of syringes), type of vascular access device (catheters instead of needles), gauge needle, arm position, syringe puncture and vacuum tube transfer, vigorous shaking or non-shaking of the sample after collection, skill of phlebotomist, characteristics of the patients related to blood vessels, etc. –– During transport: position of the tubes (horizontal or vertical), type of transport (pneumatic tube), the transport time and temperature. –– During the sample preparation: type of centrifugation, the time between blood collection and centrifugation, temperature, speed and centrifugation time, failures in the separation barrier, centrifugation of partially coagulated samples, re-centrifugation, etc. Detection of hemolysis in the laboratory is carried out by visual inspection or by automated analyzers that estimate the concentration of hemoglobin in the sample by measurement of absorbance at different wavelengths. Currently, most laboratories use automated systems. Visual detection is a subjective appraisal dependent on technician experience and hence the detection of hemolysis when hemoglobin concentration is below 0.2 g/L is difficult, due to the presence of other pigments, such as bilirubin, carotene, etc. A visually detectable minimum concentration of 0.3 g/L is being considered. The inclusion of automated hemolysis detection systems in clinical biochemical analyzers implies the removal of the subjective component, making easier the results harmonization among different laboratories. Since 2004, the group of clinical laboratories of the Institut Català de la Salut, (Catalonian Health Institute) have been evaluating the results obtained from different quality indicators, with the aim of obtaining firm quality specifications for all the processes, including the preanalytical phase processes [7–10]. To define the quality specifications of these indicators, results were chosen among the laboratory comparison results (state of the art) [11]. Regarding the indicator of hemolyzed samples, the specification obtained in this working group was 0.6%, calculated as the number of samples which invalidated
results for one or more requested tests, with regards to the number of demands. However, there was a great interlaboratory variability, ranging from 0.01% to 1.9% [10], mainly due to the following facts: the lack of standardization in hemolysis detection methods, the existence of non-homogeneous interference limits and the different characteristic of the samples collection centers. The presence of these variables, which correspond to different processes carried out in- and outside the laboratory, made it difficult to compare results for this indicator among the laboratories in the group, with the objective of obtaining quality specifications. The aim of this research was to study some factors that could have been the cause of variability in the results for the indicators. In order to develop this project, the working group set up a multi-centric study, divided into three phases: 1st phase: To study the agreement between the detection and quantification methods of hemolysis used by participant laboratories, comparing their results with those obtained by using the reference method for measuring the concentration of hemoglobin in plasma. 2nd phase: To evaluate the effect of the hemolysis degree in the results of several tests, which were measured using different analytical methods in order to establish comparable interference limits. 3rd phase: To study the influence of preanalytical variables in the hemolysis mechanism.
Materials and methods The working group was made up of 12 public laboratories: seven hospital laboratories (6 of them also with primary care) and five laboratories only with primary care belonging to the Institut Català de la Salut (ICS; Catalonian Health Institute). The mentioned laboratories are located in Catalonia, a region on the northeastern of Spain, within a radius of approximately 150 km. For the first phase of the study, quantification of the free hemoglobin concentration in serum samples was carried out, according to the adapted reference method of cyanmethemoglobin that uses potassium ferricyanide and potassium cyanide to transform hemoglobin into cyanmethemoglobin, which showed a maximum absorption at 540 nm of wavelength [12]. This method had to be adapted in order to quantify hemoglobin concentrations within the interest interval (from 0 to 10 g/L), preparing two calibration curves, one from 2 to 10 g/L and the other one from 0 to 2 g/L of hemoglobin. A spectrophotometer UV-VIS was used (Shimadzu, Kyopto, Japan) with wavelength selection at 540 nm, reactive dilution Drabkin (Sigma-Aldrich, St. Louis, MO, USA), Solution Brj35 (Sigma®) as non-ionic detergent and human lyophilized hemoglobin (Sigma®). Samples were prepared from a serum pool without apparent hemolysis. In addition a pool of whole blood samples, collected in a tube with lithium heparin, was done. Three freezing and de-freezing
Brought to you by | SUNY Buffalo Libraries Authenticated Download Date | 10/13/14 3:21 PM
Fernandez et al.: Harmonization between laboratories in hemolysis detection and interference limits 1559 cycles (at –70 °C) for the red blood cells lysis and later centrifugation at 1500 g for 10 min were done in order to obtain an hemolyzed pellet. Serial dilutions were carried out by adding this pellet to serum pool to obtain samples within the following theoretical concentration of hemoglobin: 0; 0.16; 0.30; 0.57; 0.97; 2.43; 4.85 and 6.93 g/L (target values). The hemoglobin measurement and sample dilutions were carried out in one reference laboratory of the group. The samples were collected in the reference laboratory the day before to carry out the tests and then they were taken to every laboratory. The reference laboratory location allowed that transport time to other laboratories was not longer than 1 h. For the automated detection, HI were evaluated, in triplicate, by the following analyzers: four Cobas 711 analyzers (Hoffmann-La Roche Ltd, Basel, Switzerland), three AU 5400 (Beckman Coulter Inc, Fullerton, CA, USA), one Advia 2400 analyzer (Siemens Healthcare Diagnostics Inc, Deerfield, IL, USA), one DXC 800 (Beckman), one Synchron LXi725 (Beckman), one Cobas 6000 (Roche), one Modular DP (Roche), one Vista (Siemens). Different ways to express HI detection for these analyzers were summarized in Table 1. To evaluate the effect of the hemolysis degree in the results for different analytes (second phase), all the samples were analyzed in triplicate under same operating conditions for the following tests: alanine aminotransferase: (ALT), aspartate aminotransferase (AST), creatine kinase (CK), cholesterol (COL), phosphorus (P), alkaline phosphatase (ALP), iron (FE), glucose (GLU), γ-glutamyl-transferase (GGT), potassium (K), lactate dehydrogenase (LD), total protein (TP), triglycerides (TG), using all the mentioned analyzers. The analytical methods used to measure the different analytes, are shown in Table 2. The average of the three carried out determinations, was calculated for each hemoglobin standard and for each test. For each test, the obtained value in the standard of hemoglobin concentration 0 g/L was taken as baseline and the differences in percentage were calculated between the results obtained in each standard in comparison to the baseline. The cut-off point to define hemolysis interference for a specific test in each laboratory was established, when the percentage surpassed the bias derived from biological variation database [13, 14].
For each analyzer the interference limit was established as the median of the cut-off points obtained in each laboratory. To study the most influential variables in hemolysis production (third phase), the percentage of K result rejection in regard to those requested results, was calculated, considering rejection when the hemolysis exceeded the interference limit obtained in the second phase of the study for this test. For 1 month, in 324 primary care blood collection centers, where the samples transport (in refrigerators or at room temperature) was carried out by road to seven of the participant laboratories, the following data were collected as critical points: 1. Extraction system, (syringe or vacuum system) taking into account the provider and the sample volume for each tube; BD Vacutainer blood collection tubes (3.5 mL and 8.5 mL) (Becton Dickinson, NJ, USA) and Vacuette blood collection tubes (4 mL and 10 mL) (Vacuette, Greiner Bio-one, Kremsmuenster, Austria). 2. Centrifugation site (in the blood collection center or in the laboratory) and centrifugation conditions. 3. Transport time. 4. Temperature.
Statistical analysis Weighted κ coefficient (http://www.vassarstats.net/kappa.html) and intra-class correlation coefficient (ICC) were used for comparability of automated detection of hemolysis in regard to reference method. A weighted κ coefficient above 0.8 and an ICC above 0.75 were considered as a good agreement [15]. A forward multiple regression analysis has been carried out to assess the effect of such preanalytical characteristics upon the rejection percentage, so that the equation turns to be defined by those variables automatically included and their associated coefficients. The odds ratio was calculated – for each level – dividing the percentage of rejected specimens by the baseline percentage of rejected specimens (the baseline level for each variable was the level occurring most often). The transport time variable was divided into several intervals and the odds ratio was calculated with regards to a time interval 0–15 min.
Table 1 Different ways to express hemolytic indexes for studied analyzers.
Result
Beckman
Synchron Lxi725–DXC800
Qualitative 0 Not detected 1 Hb = (0–0.5) 2 Hb = (0.5–1) 3 Hb = (1–1.5) 4 Hb = (1.5–2) 5 Hb = (2–2.5) 6 Hb = (2.5–3) 7 Hb = (3–3.5) 8 Hb = (3.5–4) 9 Hb = (4–4.5) 10 Hb = (4.5–5)
Equivalence range hemoglobin, g/L
Siemens
AU 5400
Vista
Qualitative
Qualitative
0 (Hb ≤ 0.5) + (0.5 5 > 2– ≤ 3 > 0.5– ≤ 2 1.90 ± 0.01 1.83 ± 0.01 1.88 ± 0.02 1.80 ± 0.00 1.91 ± 0.02 1.74 ± 0.12 1.82 ± 0.02
> 5 > 5 > 5 > 5 > 5 > 3– ≤ 5 3.76 ± 0.01 3.57 ± 0.02 3.67 ± 0.01 3.52 ± 0.03 3.70 ± 0.04 3.37 ± 0.02 3.54 ± 0.02
> 5 > 5 ABNa ABNa ABNa > 3– ≤ 5 5.18 ± 0.02 4.99 ± 0.04 5.13 ± 0.03 4.91 ± 0.03 5.19 ± 0.04 4.73 ± 0.05 4.96 ± 0.02
a ABN, sample flagged as abnormal by the AU 5400 reagent; bHI qualitatives, In all cases the three observations were equivalent and expressed as an interval; cHI quantitatives, for the different analyzers expressed as mean ± SD. HI, hemolytic indexes.
Brought to you by | SUNY Buffalo Libraries Authenticated Download Date | 10/13/14 3:21 PM
1562 Fernandez et al.: Harmonization between laboratories in hemolysis detection and interference limits Table 4 κ of Cohen index and coefficients of intra-class correlation.
Beckman Synchron Lxi725 –DXC800
κ CI 95%
0.7895 0.6899–0.8891
AU 5400
0.8326 0.7278–0.9374
Siemens Vista
0.8209 0.6587–0.9831
Discussion There is a great variability in the percentage of hemolyzed samples in different laboratories, in a range between 0.05% and 3.3% according to different authors. The variation sources could be due to the different ways of obtaining and preparing samples, the different methods used by laboratories to detect and quantify hemolysis, the different criteria to set rejection interference limits for hemolyzed samples, as well as the different formulas used to calculate the indicator, which could be related to the number of blood samples, total samples, serum tubes or requests [3]. The final aim of our study was to evaluate preanalytical variables with most influential in the hemolysis production, after setting homogeneous interference limits for determined tests in all participant laboratories, as well as to establish the same rejection criteria and calculation formula. The reference method adapted for measurement of the hemoglobin concentration allowed us to quantify concentrations below 1 g/L, with an acceptable analytical performance for the study objective.
Siemens
CCI CI 95%
0.982 0.910–0.996
Advia
Roche Cobas-Modular
0.973 0.863–0.995
Visual hemolysis detection made difficult to carry out comparative studies among laboratories, due to the fact that this process was a low sensitive and subjective assessment which depends on the laboratory staff. In this sense, Glick et al. [17] and Simundic et al. [18] have reported disagreement among concentrations assigned visually and the concentration of real hemoglobin of a series of patterns and automated HI detection, respectively. There were a few studies comparing hemoglobin concentrations measured by the reference method to the HI obtained in different analyzers. Our results showed a high agreement for the most studied analyzers, similar to the results obtained by Lippi et al. [19]. The highest agreement regarding to reference method was obtained in those analyzers which give quantitative information about HI, although the results were undervalued, the homogeneity was allowed for all the analytical systems to set up a common indicator. According to our results, for some tests such as LDH, K and P (the majority released in the hemolysis process) homogeneous interference limits were obtained
Table 5 Proposed hemolysis interference limit (median per group of analyzers), (A) versus limit recommended by the manufacturer (B). Test
Hemolysis interference limit, g/L Advia 2400 (Siemens) A
ALT ASTa CKa COL P FAL FEa GLUa GGTa K LDa PTa TG a
2.4 0.6 2.4 > 6.9 2.4 2.4 4.8 2.4 6.9 0.6 0.16 6.9 6.9
B 5 5 5 7.5 5.25 5 na 5.25 5 na na 5 5.25
Synchron LXi725–DXC 800 (Beckman)
AU 5400 (Beckman)
A
B
A
1.7 0.3 1.7 > 6.9 3.6 4.8 0.6 > 6.9 3.6 0.6 0.16 3.6 3.6
0.5 0.5 0.5 5 1.5 5 0.5 5 2.5 0.5 0.5 5 5
2.4 0.6 4.8 2.4 2.4 2.4 6.9 > 6.9 6.9 0.6 0.16 2.4 6.9
B 5 na 1 5 3.5 4.5 1 na 5 na na 3 5
Cobas 711,6000 (Roche) A
Test with different analytical methods according to the analyzers. na, not available.
a
Brought to you by | SUNY Buffalo Libraries Authenticated Download Date | 10/13/14 3:21 PM
2.4 0.4 2.4 > 6.9 2.4 4.8 2.4 > 6.9 4.8 0.6 0.16 2.4 4.8
B 2 0.4 2 7 3 2 2 10 2 na 0.15 10 7
Vista (Siemens) A
B
4.8 0.2 4.8 4.8 2.4 4.8 2.4 0.9 > 6.9 1 0.16 0.9 > 6.9
> 10 0.25–0.5 5–10 > 10 5–10 > 10 0.25–0.5 > 10 5–10 0.25 0.1–0.25 > 10 > 10
Fernandez et al.: Harmonization between laboratories in hemolysis detection and interference limits 1563
AST. % Dif f erences between the results obtained vs baseline (22.4 U/L)
ALT. % Dif f erences between the results obtained vs baseline (18.1 U/L) 175.0
35.0 30.0
Beckman (Synchron Lxi 725/ DXC 800)
150.0
25.0
Beckman AU
125.0
20.0
Siemens (Vista)
15.0
Siemens (Advia 2400)
10.0
Roche (Cobas/Modular)
5.0
Optimal Spec (5.7%)
0.0 0.16
0.3
0.57
0.97
2.43
4.85
Beckman (Synchron Lxi 725/ DXC 800) Beckman AU
100.0
Siemens (Vista)
75.0
Siemens (Advia 2400)
50.0
Roche (Cobas/Modular)
25.0
Desirable Spec (5.4%)
0.0 0.16
6.93
0.3
0.57
0.97
2.43
4.85
6.93
Hemoglobin added, g/L
Hemoglobin added, g/L CK. % Dif f erences between the results obtained vs baseline (137.5 U/L)
Cholesterol. % Dif f erences between the results obtained vs baseline (209.4 mg/dL)
35.0
15.00
30.0
Beckman (Synchron Lxi 725/ DXC 800)
25.0
Beckman AU
20.0
Beckman (Synchron Lxi 725/ DXC 800)
10.00
Beckman AU
5.00
Siemens (Vista)
Siemens (Vista)
15.0
Siemens (Advia 2400)
10.0
Roche (Cobas/Modular)
5.0
0.00
Siemens (Advia 2400)
-5.00
Roche (Cobas/Modular) Desirable Spec (4%)
-10.00
Optimal Spec (5.8%)
0.0 0.16
0.3
0.57
0.97
2.43
4.85
-15.00
6.93
0.16
0.3
Hemoglobin added, g/L
0.57
0.97
2.43
4.85
6.93
Hemoglobin added, g/L
Alkaline Phosphatase. % Dif f erences between the results obtained vs baseline (69.1 U/L)
Phosphate. % Dif f erences between the results obtained vs baseline (4.5 mg/dL)
15.00
10.00 Beckman (Synchron Lxi 725/ DXC 800)
10.00
Beckman AU
5.00
Beckman (Synchron Lxi 725/ DXC 800)
0.00
Beckman AU
-5.00 Siemens (Vista)
5.00
Siemens (Vista)
-10.00 Siemens (Advia 2400)
0.00
-5.00
-15.00
Siemens (Advia 2400)
Roche (Cobas/Modular)
-20.00
Roche (Cobas/Modular)
Desirable Spec (3.2%)
-25.00
Desirable Spec (6.4%)
-30.00 0.16
0.3
0.57 0.97 2.43 4.85 Hemoglobin added, g/L
6.93
0.16
0.3
0.57 0.97 2.43 4.85 Hemoglobin added, g/L
6.93
(Figure 1 Continued)
Brought to you by | SUNY Buffalo Libraries Authenticated Download Date | 10/13/14 3:21 PM
1564 Fernandez et al.: Harmonization between laboratories in hemolysis detection and interference limits Glucose. % Dif f erences between the results obtained vs baseline (99.4 mg/dL )
Iron. % Dif f erences between the results obtained vs baseline (98.4 µg/dL ) 10.00
35.00 30.00
Beckman (Synchron Lxi 725/ DXC 800)
25.00
Beckman AU
20.00 15.00
Siemens (Vista)
10.00
Siemens (Advia 2400)
Beckman (Synchron Lxi 725/ DXC 800)
5.00
Beckman AU
0.00 Siemens (Vista)
-5.00
5.00
Siemens (Advia 2400)
-10.00
Roche (Cobas/Modular)
0.00
Optimal Spec (4.4%)
-5.00
Roche (Cobas/Modular)
-15.00
Desirable Spec (2.2%)
-20.00
-10.00 0.16
0.3
0.57
0.97
2.43
4.85
0.16
6.93
0.3
Hemoglobin added, g/L
0.57
0.97
2.43
4.85
6.93
Hemoglobin added, g/L
GGT. % Dif f erences between the results obtained vs baseline (28.0 U/L)
Potassium. % Dif f erences between the results obtained vs baseline (4.3 mmol/L) 30.00
15.00 10.00
Beckman (Synchron Lxi 725/ DXC 800)
5.00
Beckman AU
0.00
Beckman (Synchron Lxi 725/ DXC 800)
25.00
Beckman AU
20.00
Siemens (Vista)
Siemens (Vista)
15.00
-5.00 -10.00
Siemens (Advia 2400)
-15.00
Roche (Cobas/Modular)
-20.00
Optimal Spec (5.4%)
-25.00
0.16
0.3
0.57
0.97
2.43
4.85
Siemens (Advia 2400)
10.00
5.00
Roche (Cobas/Modular)
0.00
Minimal Spec (2.8%)
0.16
6.93
0.3
0.57
0.97
2.43
4.85
6.93
Hemoglobin added, g/L
Hemoglobin added, g/L
T. Protein. % Dif f erences between the results obtained vs baseline (7.5 g/L)
LD. % Dif f erences between the results obtained vs baseline (330 U/L) 15.00
300.00 Beckman (Synchron Lxi 725/ DXC 800)
250.00
Beckman (Synchron Lxi 725/ DXC 800)
10.00
Beckman AU
Beckman AU
200.00
Siemens (Vista)
5.00 Siemens (Vista)
150.00
100.00
Siemens (Advia 2400)
50.00
Roche (Cobas/Modular) Desirable Spec (4.3%)
0.00
0.16
0.3
0.57 0.97 2.43 4.85 Hemoglobin added, g/L
6.93
Siemens (Advia 2400)
0.00
Roche (Cobas/Modular)
-5.00
Minimal Spec (1.8%)
-10.00
0.16
0.3
0.57 0.97 2.43 4.85 Hemoglobin added, g/L
6.93
(Figure 1 Continued)
Brought to you by | SUNY Buffalo Libraries Authenticated Download Date | 10/13/14 3:21 PM
Fernandez et al.: Harmonization between laboratories in hemolysis detection and interference limits 1565
15.00
Triglyceride. % Differences between the results obtained vs baseline (133.7 mg/dL)
Beckman (Synchron Lxi 725/ DXC 800)
10.00 5.00
Beckman AU
0.00
Siemens (Vista)
-5.00
Siemens (Advia 2400)
-10.00
Roche (Cobas/Modular)
-15.00
Optimal Spec (5.3%)
-20.00 0.16
0.3
0.57 0.97 2.43 4.85 Hemoglobin added, g/L
6.93
Figure 1 Percentage difference for each standard in comparison to the baseline.
among the different analyzers. However, for the rest of the analyzed tests, the hemolysis interference cut-off point depended on the method and/or analyzer studied (Table 5). There were discordant results among analyzers for some parameters, e.g., Fe, with an interference limit between 0.6 and 6.9 g/L or GLU with a limit between 0.9 and > 6.9 g/L depending on the analyzer. When these results were compared with those published in different studies, discrepancies were also observed in interference limits for different tests [14, 20]. As has been mentioned previously in the Introduction, hemoglobin concentrations below 0.3 g/L are not easily detected by visual inspection; for this reason, hemolysis detection would be advisable in the analyzers for those analytes (LDH, AST and K), which showed an interference
limit below the minimum concentration visually detected [14]. In the same way, it is advisable that all analyzers detect the previous concentrations. According to our results, it has been observed that analyzers Beckman AU 5400, DXC800 and Synchron LXi725, which carried out a qualitative assessment of the HI, were not able to detect hemoglobin concentrations below 0.5 or 1 g/L and, therefore, significant interferences were not detected for some tests. As previously exposed, the implementation of HI in analyzers is advisable, because it also has been recently demonstrated by Lippi et al. [21] that systematical measurement of HI does not impair instrument efficiency. There is not universal consensus to set up the maximum allowable bias due to hemolysis [22]. The allowable error for interference based on the biological variation of the analyte [23] defines an interference clinically significant, when the result in the presence of the interferent differs more than 1.96 × (CVa2+CVw2)½ from the result without the interferent, where CVa is the analytical coefficient of variation and CVw the within-subject biological variation. The IFCC Working Group on Allowable Errors for Traceable Results [24] has stated that this bias should not exceed a definite part of total error uncertainty (total error budget) of a measurement. The considered criterion in this research as well as, Lippi et al. [14] to establish the thresholds for hemolysis interference for each measured analyte, was defined on the basis of comparison of desirable bias derived from biological variation [13]. This criterion has narrowed the analytical variation allowable only to the biological variation, and we believe that is a way to detect the interference for hemolysis with higher sensitive.
Table 6 Rejection and odds percentages calculated for the considered preanalytical variables. Preanalytical variables
Centers, n
Rejections, n
Serum tubes, n
Rejections, %
Odds (L.Inf-L.Sup CI 95%)
Transport time 0–15 min Transport time 16–60 min Transport time 61–120 min Transport time 120–150 min Transport at room temperature Transport at cooled temperature Centrifugation in laboratory Centrifugation in blood collection center Tubes vacuette Tubes BD-vacutainer Tubes volume 3.5 and 4 mL Tubes volume 8, 8.5 and 9 mL Vacuum extraction system Syringe extraction system
24 153 126 18 155 164 312 12 90 234 221 103 319 3
210 802 536 28 792 708 1552 21 610 963 1109 464 1543 22
11,154 32,608 19,832 982 39,019 20,658 62,972 1775 24,351 40,396 48,575 16,172 63,968 416
1.88 2.46 2.70 2.85 2.03 3.43 2.46 1.18 2.51 2.38 2.28 2.87 2.41 5.29
a
Baseline variable.
a
Brought to you by | SUNY Buffalo Libraries Authenticated Download Date | 10/13/14 3:21 PM
1.306 (1.124–1.518) 1.436 (1.289–1.599) 1.514 (1.041–2.202) a
1.688 (1.528–1.866) a
0.480 (0.313–0.736) 1.051 (0.950–1.162) a a
1.257(1.129–1.398) a
2.192 (1.455–3.302)
1566 Fernandez et al.: Harmonization between laboratories in hemolysis detection and interference limits In the in vitro samples used in our study, the influent variation was due to analytical inaccuracy in addition to hemolysis interference. The influence of biological variation was avoided by calculating the percentual difference among results of each standard with regards to basal value. Carrying out determinations in the same analytical batch kept the variability due to analytical bias as low as possible. Finally, carrying out serial determinations, the analytical imprecision was kept in a low level. For this reason the percentage of variation regarding baseline will be due to hemolysis interference. Discrepancies were also observed between the interference limits recommended by manufacturers and those obtained in our study. It is worth noting, the difference observed in cut-off used by different manufacturers. As was shown in Table 5 for phospate, the interference limit range among manufacturers was 1.5–10 g/L, however, the interference limit range for hemolysis in our results for different equipment ranged from 2.4 to 3.6 g/L. These discrepancies were probably due to differences in the criteria chosen to establish the cut-off point, which makes comparison difficult. In our case the cut-off was established when the difference (in percentage) between the result of the test on each standard and the baseline (HI: 0 g/L) exceeded the systematic error based on biological variation. In this sense manufacturers should make an effort to supply the complete data for the interference study (interferograms), being shown the type of study and the methodology utilized for the hemolysis quantification, as well as providing indexes which can discriminate free hemoglobin concentrations below 0.5 g/L. According to our results, it would be advisable for each laboratory to verify the information supplied by providers with regards hemolysis interference and to set up interference limits for those biochemical tests susceptible to be interfered, according to their own quality specifications. So far, no manufacturer has supplied calibrators and controls for serum indexes (hemolytic, icteric or lipemic), making difficult to control these methods. The inclusion of internal and external quality controls for these automated detection methods of serum index would mean an advance in standardization and results inter-comparison. According to our results, COL, GLU, GGT and TGs have an interference limit for hemolysis higher than 6.9 g/L, in some studied analyzers. At the moment of carrying out the study, a high volume of sample was requested (from Blood Bank) in order to supply aliquots of all the standards to the 12 participant laboratories of the study. Despite the recommended maximum concentration of hemoglobin to be tested to establish the
hemolysis interference is 10 g/L [25], due to the percentage of hemolyzed samples with a high hemolysis degree ( > 6.9 g/L) received in the laboratory is less than the total percentage of hemolyzed samples, a new concentration limit of hemoglobin in 6.9 g/L was suggested. As an example, the percentage of samples with a hemolysis degree higher than 6 g/L was calculated for 3 months in one of the participant laboratory groups and a value of 0.02% of a total 81,390 hemolyzed samples was obtained. The last phase of our study was approached once homogeneous interference limits had been set up for each test in every involved laboratory. In this phase, rejection percentages obtained for K analyte were calculated in order to determine the preanalytical variable with the most influence on the in vitro hemolysis production. Of all the variables which could influence hemolysis of samples, those considered most relevant and which could be collected without error by all the participant laboratories were selected (Table 6). In most cases centrifugation was carried out in the laboratory and in some further centers the centrifugation was carried out in the same collection center (centrifugation in origin), but in this case the samples were sent to the laboratory after centrifugation. The hemolysis rejections percentage has been increased significantly depending on transport time from 1.88%, if transport time was
