Interferograms plotted with reference change value (RCV) may facilitate the management of hemolyzed samples

Background The European Federation of Clinical Chemistry and Laboratory Medicine (EFLM) Working Group for Preanalytical Phase (WG-PRE) have recommended an algorithm based on the reference change value (RCV) to evaluate hemolysis. We utilized this algorithm to analyze hemolysis-sensitive parameters. Methods Two tubes of blood were collected from each of the 10 participants, one of which was subjected to mechanical trauma while the other was centrifuged directly. Subsequently, the samples were diluted with the participant's hemolyzed sample to obtain the desired hemoglobin concentrations (0, 1, 2, 4, 6, 8, and 10 g/L). ALT, AST, K, LDH, T. Bil tests were performed using Beckman Coulter AU680 analyzer. The analytical and clinical cut-offs were based on the biological variation for the allowable imprecision and RCV. The algorithms could report the values directly below the analytical cut-off or those between the analytical and clinical cut-offs with comments. If the change was above the clinical cut-off, the test was rejected. The linear regression was used for interferograms, and the hemoglobin concentrations corresponding to cut-offs were calculated via the interferograms. Results The RCV was calculated as 29.6% for ALT. Therefore, ALT should be rejected in samples containing >5.9 g/L hemoglobin. The RCVs for AST, K, LDH, and T. Bil were calculated as 27.9%, 12.1%, 19.2%, and 61.2%, while the samples' hemoglobin concentrations for test rejection were 0.8, 1.6, 0.5, and 2.2 g/L, respectively. Conclusions Algorithms prepared with RCV could provide evidence-based results and objectively manage hemolyzed samples.


Summary
Background: The European Federation of Clinical Chemistry and Laboratory Medicine (EFLM) Working Group for Preanalytical Phase (WG-PRE) have recommended an algorithm based on the reference change value (RCV) to evaluate hemolysis. We utilized this algorithm to analyze hemolysis-sensitive parameters. Methods: Two tubes of blood were collected from each of the 10 participants, one of which was subjected to mechanical trauma while the other was centrifuged directly. Subsequently, the samples were diluted with the participant's hemolyzed sample to obtain the desired hemoglobin concentrations (0, 1, 2, 4, 6, 8, and 10 g/L). ALT, AST, K, LDH, T. Bil tests were performed using Beckman Coulter AU680 analyzer. The analytical and clinical cut-offs were based on the biological variation for the allowable imprecision and RCV. The algorithms could report the values directly below the analytical cut-off or those between the analytical and clinical cut-offs with comments. If the change was above the clinical cut-off, the test was rejected. The linear regression was used for interferograms, and the hemoglobin concentrations corresponding to cut-offs were calculated via the interferograms. Results: The RCV was calculated as 29.6% for ALT. Therefore, ALT should be rejected in samples containing >5.9 g/L hemoglobin. The RCVs for AST, K, LDH, and T. Bil were calculated as 27.9%, 12.1%, 19.2%, and 61.2%, while the samples' hemoglobin concentrations for test rejection were 0.8, 1.6, 0.5, and 2.2 g/L, respectively.

Introduction
The International Organization for Standardization (ISO) has established that laboratory errors may arise at every point related to the laboratory in the process, from the test request to diagnosis and treatments interrelated to test results (1). This comprehensive definition showed that the laboratory's responsibilities are not only restricted to analyze the test and report its results. Nowadays, when laboratory results are considered crucial in clinical decisions, laboratories have a significant role, directly or indirectly, in the diagnosis and treatment of patients in many aspects (2). Thus, laboratory errors should be handled with a holistic approach, evaluated, and fixed (3).
Most of the errors evident in laboratories arise from the preanalytical process (4). Notably, the most common problem in this process is that of inappropriate samples (5). Among inappropriate samples, hemolyzed samples are the most common. Evaluation of all samples sent to the laboratory in a previous study, revealed that 2.2% of them were hemolyzed samples (6). Among samples accepted from emergency departments, up to 30% have been reported to be hemolyzed (7). Hemolysis can be in vivo and in vitro. In vivo hemolysis occurs during various diseases or treatments and constitutes only 2% of all hemolyzed samples accepted to the laboratory (8). In vitro hemolysis is almost the sole cause of hemolyzed samples encountered in laboratories and is caused by the following: blood collection process, sample handling process, sample processing process, and individual differences (9).
Hemolysis can be evaluated using two methods. The first is the classical method based on the technician's visual assessment. The visual evaluation method is discouraged because of its subjectivity and low reproducibility. Even if the color indicator charts with varying degrees of hemolysis, sample photographs are used for evaluation; this method's reliability and sensitivity are relatively low (10,11). The second method uses hemolysis index (H-index), a tool that enables estimating the sample free hemoglobin (Hb) concentration using automated systems. The H-index can report the degree of hemolysis to both the laboratory specialist and physician. Therefore, it can enable the evaluation of the test results affected by hemolysis in the analyzed sample. Automated systems are reportedly more reliable than visual evaluation in determining hemolysis, and therefore, it is recommended for use instead of the visual method (12). Currently, the management of hemolyzed samples involves applying four options according to the hemolysis level, as follows: analysis of the sample followed by reporting the result; reporting the result and its interpretation; rejection based on the test; thorough sample rejection (13). For correct implementation, the serum index should be determined based on the test for the management of hemolysis.
The C56-A guideline of the Clinical and Laboratory Standards Institute (CLSI) recommends that interference studies should first be carried out by in vitro diagnostic (IVD) manufacturers and that the laboratories should design, as well as use, their algorithms. It also states that the reference change value (RCV) can be used as the allowable total error criterion for interference studies (14). Fraser stated that when the interfering effect of the substance on the test result exceeds the RCV, it could change the actual level of the measured analytes, with a clinically significant difference (15). The entire procedure was presented in an opinion letter that was recommended for clinical biochemistry tests published by the European Federation of Clinical Chemistry and Laboratory Medicine (EFLM) Working Group for Preanalytical Phase (WG-PRE) in 2018 (16). The EFLM WG-PRE has proposed an algorithm in line with this data to evaluate hemolysis interference for clinical chemistry tests. Interferograms were created by adding the allowable analytical coefficient of variation (CV) values and RCV as evaluation criteria after graphs were drawn by calculating the percentage change according to the Hb concentration for each test. Recent studies recommend a new algorithm that allows the interpretation of the test result through interferograms (12).
In this study, we aimed to develop algorithms to be used in the alanine aminotransferase (ALT), aspartate aminotransferase (AST), lactate dehydrogenase (LDH), potassium (K), and total bilirubin (T. Bil) tests for our laboratory, with respect to these suggestions, and to compare the data obtained with the manufacturer's statements.

Materials and Methods
This study was performed at the Medical Biochemistry Laboratory of the Gaziosmanpasa Training and Research Hospital, with the approval of the local ethics committee (decision date & number: April 10, 2019-56), in accordance with the Declaration of Helsinki. The study was explained to all volunteers and then informed consent was obtained.

Study design and sample preparation procedure
The sample size was determined to be 10 people, as per the CLSI EP 07A2 guideline. Acute or chronic disease, regular medication use, bleeding disorder, and pregnancy were selected as the exclusion criteria. Two tubes of blood were collected from 10 apparently healthy volunteers aged between 18 and 50 years. A single experienced phlebotomist was appointed to perform blood collection, to avoid hemolysis caused by phlebotomy. The samples were collected in the collection tubes (BD Barricor 5.0 mL, 13x100 mm, Becton, Dickenson and Company, NJ, USA) using the blood collection device (BD Vacutainer Holder, Becton, Dickenson and Company, NJ, USA) equipped with a needle (BD Vacutainer Eclipse 21G, BD-Belliver Industrial Estate, Plymouth PL6 7BP, UK).
Conclusions: Algorithms prepared with RCV could provide evidence-based results and objectively manage hemolyzed samples.
Klju~ne re~i: hemoliza, interferencija, preanaliti~ka faza, laboratorijske gre{ke, promene referntnih vrednosti One of the paired tubes was centrifuged without any intervention to obtain the hemolysis-free samples. The blood from the other tube was passed through the needle 10 times using an injector to achieve mechanical hemolysis (12). The mechanical trauma method was preferred to obtain samples similar to hemolyzed samples sent to the laboratory. This method ensured that the hemolyzed sample contains leukocytes and thrombocytes (13). Subsequently, the tubes were centrifuged at 2000 × g for 10 min following the manufacturer's recommendation.
Free Hb concentrations were measured in the separated plasma samples by the spectrophotometric method using the auto hematology analyzer (Mindray BC 6800, Shenzhen Mindray Bio-Medical Electronics Co., China; imprecision of Hb: 1.0%). It was confirmed that one of the samples had an Hb concentration of 0 g/L, while the other had a Hb concentration >10 g/L. To obtain samples with the desired free Hb concentrations (1, 2, 4, 6, 8, and 10 g/L), the hemolyzed samples were diluted with the hemolysis-free sample of the same volunteer. The samples were prepared individually for each volunteer, thus preventing dilution bias and inter-individual differences from affecting the values obtained. After obtaining the diluted samples, each sample was analyzed duplicate on the Mindray BC 6800 analyzer to confirm the desired Hb concentration. The process for sample preparation is shown in Figure 1. Figure 1 Preparation of hemolyzed samples. The samples were prepared individually for each volunteer (n=10). As a result, samples with the desired Hb concentrations (0, 1, 2, 4, 6, 8, and 10 g/L) were procured. Hb was expressed as g/L.
Two tubes of blood were collected from each volunteer.
One of the tubes was centrifuged without any intervention.
The blood in the other tube was passed through needles (13 mm, 26 G) 10 times, and centrifuged.
A specimen with 10 g/L free Hb concentrations was obtained by diluting these samples with each other in various proportions for each volunteer. In the technical sheets of reagents, it has been reported that hemolyzed samples should not be used for AST, K, and LDH tests. No Hb concentration, H-index limit, or bias of interference has been specified. It has been reported that the interference could be up to 10% in samples with 5 g/L free Hb for ALT and 0.45 g/L free Hb for T. Bil. H-index values are expressed qualitatively on the Beckmann AU680 analyzer. The cut-off degrees and the corresponding approximate Hb concentration ranges are presented in Table I.

Statistical analysis
The percentage difference was calculated for the samples from each patient. Next, the mean percentage differences were calculated for each Hb concentration using the percentage differences for 10 different volunteers. The Shapiro-Wilk test was used to evaluate whether the values were normally distributed. All values in our study were distributed normally, and the results are presented as the mean ± SD. Subsequently, interferograms were plotted for the evaluated tests. The scatter plots and linear regression method were used when creating interferograms, and the Hb concentrations corresponding to analytical and clinical cut-offs were calculated using the regression equation of the graphs. Analytical cut-offs were determined according to the desirable allowable CV (analytically acceptable imprecision (I %)) values of the biological variation (BV) databases. Clinical cut-offs were determined according to the RCV. Intra-individual CV (CV I ) for tests was obtained from the EFLM BV database and Westgard Desirable BV database (17,18). The Hb concentration for sample rejection was determined to be 10 g/L following the recommendation of EFLM (16). The formulas for the calculation are as follows:

Results
According to the specified tests using samples with determined Hb concentrations, mean percentage differ-56 Uçar et al.: Interferograms plotted with RCV to manage hemolysis    Report the test result with the note »Possible affected by hemolysis«. Ask for a new sample.
Reject the sample due to hemolysis and ask for a new sample.
ences (%) between the samples are presented in Table II.
The interferograms of the tests are shown in Figure 2. The desirable allowable impression (I %) -analytical cut-off for ALT was obtained as 5.0%. The RCV -clinical cut-off was calculated as 29.6% (Table III). Hb concentrations corresponding to these values were calculated as 0.8 and 5.9 g/L based on the regression equation (Table III). Accordingly, for samples with Hb concentration <0.8 g/L, ALT results could be reported directly; with 0.8-5.9 g/L, ALT results could be reported with interpretation (Table III).
Since the percentage change values of ALT in samples with Hb >5.9 g/L were greater than the RCV, it was deemed appropriate for rejection (Table III) Table III). The Hb concentration cut-offs obtained from the regression equation for the four tests were found to be 0.2, 0.8 g/L; 0.2, 1.6 g/L; 0.3, 0.5 g/L, and 0.5, 2.2 g/L; respectively ( Table III). The H-index measurement results in each hemolyzed sample were observed in agreement with the manufacturer's H-index cut-offs. However, it was observed that the ratio of change in analyte results due to hemolysis did not fully comply with the manufacturer's statements.
The CV A , CV I , analytical cut-off (I) and clinical cut-off values (RCV), regression equations (95% CI), coefficient of determinations (R 2 ), and the corresponding Hb levels to cut-off values are presented in Table III. A simple presentation of the algorithms is shown in Figure 3.

Discussion
In our study, the hemolysis interference was evaluated for five clinical biochemistry parameters, based on the BV desirable I% (analytical cut-off) and RCV% (clinical cut-off). Accordingly, to determine the extent to which the tests were affected by hemolysis, LDH, AST, K, T. Bil, and ALT tests were performed, with respect to Hb concentration ranging from low to high.
For LDH, if the Hb concentration of the sample was <0.3 g/L, the test result could be reported directly; moreover, it could be reported with comment for 0.3-0.5 g/L Hb, and the test had to be rejected for >0.5 g/L Hb. For AST and K, the corresponding Hb values (in g/L) were found to be <0.2, 0.2-0.8, and >0.8, and <0.2, 0.2-1.6, and >1.6, respectively ( Figure 3). As per the manufacturer's recommendation, these three tests should be rejected on the hemolyzed samples. As the initial cut-off value determined by the manufacturer for the H-index is 0.5 g/L, it can be assumed that the hemolyzed sample was meant for samples containing 0.5 g/L Hb. In another study, the upper limit of Hb reference value for hemolysis-free plasma samples was found to be 0.13 and 0.10 g/L in two different biochemistry analyzers (19). Although no information is available in the literature for our device, these values may be considered valid for our study, since it involved working with plasma samples containing lithium heparin. In fact, the Hb concentrations corresponding to the analytical cut-offs we calculated in our study for LDH, AST, and K seem close to these values. Besides, the tests most affected by hemolysis were those of LDH, AST, and K. However, our findings -particularly those associated with K -do not seem entirely consistent with the manufacturer's statement. It has been observed that K can be analyzed on samples containing up to 1.6 g/L Hb.
When the Hb concentration of the sample for T. Bil was <0.5 g/L, the result could be reported directly. It was observed that for 0.5-2.2. g/L Hb, the result could be reported with comment and that it should be rejected for >2.2 g/L Hb. For ALT, these Hb values were found to be <0.8, 0.9-5.9, and >5.9, respectively ( Figure 3). The manufacturer stated that the hemolysis interference could be less than 10% in samples containing Hb up to 0.45 g/L for T. Bil and 5 g/L for ALT. Although this information is consistent with our findings, it is insufficient to manage samples containing 0.45 and 5 g/L in the T. Bil and ALT, respectively. According to our findings, samples containing up to 2.2 g/L and 5.9 g/L could be analyzed for T. Bil and ALT, respectively, and the result could be reported with the interpretation.
Perovic and Dolvic (20) evaluated the hemolysis levels of 25 clinical biochemistry parameters using RCV and compared the results with the manufacturer's statements. Similar to our study, the Beckman Coulter AU480 clinical chemistry analyzer and reagents from the same manufacturer were used in this study. It was observed that the findings within the reference range obtained in this study were consistent with our findings, and in the same study, it was stated that the manufacturer's statements were insufficient for ALT tests, among the tests evaluated. Studies on other devices and reagents have also obtained results that do not comply with the manufacturer's declarations (21)(22)(23)(24). On the other hand, the general approach assumes the significant change affecting the result as ± 10% and presents the cut-off values accordingly. However, this ratio is far from being a proper criterion for every test (24). Using RCV or other criteria instead of ± 10% change value as a standard for each test can also support their flexibility (24,25). Additionally, rejection rates may decrease by accepting RCV as a cut-off for test rejection (23). Based on these data, the use of RCV and these algorithms by manufacturers while performing interference studies will make the H-Index more beneficial.
Moreover, several problems have been observed with the routine use of the H-index. EFLM WG-PRE had a call for IVD manufacturers to provide more clarity, with respect to serum indexes, in 2018 (26). One problem is that the Hindex is reported as degrees in some devices and free Hb concentrations in other devices. EFLM WG-PRE recommended a harmonization in reporting results via the use of a common unit, free Hb (g/L). Another problem is that manufacturers do not adequately report the interference specifications of the kits according to their H-index. For laboratories, inadequate information could be a significant problem in the management of hemolyzed samples. This situation was also evaluated in a large survey study conducted by EFLM WG-PRE, with the participation of 1405 laboratories in 37 European countries. It has been reported that many laboratories indicate heterogeneity of data on interference as the reason for avoiding serum indexes, and 67% of laboratories using serum indexes use the cut-off values recommended by the manufacturer without verification (27,28). On the other hand, EFLM WG-PRE assumes that IVD producers did not fully comply with CLSI guidelines while performing interference studies; therefore, verification studies should be carried out by the laboratories (28). Furthermore, internal, and external quality control evaluations for H-index are recommended, given the critical importance of H-index results in the evaluation of other test results (26). It is thought that the approaches in which pre-