The automated processing algorithm to correct the test result of serum neuron‐specific enolase affected by specimen hemolysis

Abstract Introduction Serum neuron‐specific enolase (NSE) is an important tumor marker for small cell lung cancer and neuroblastoma. However, the test of serum NSE compromised by specimen hemolysis is presented as a falsely higher result, which seriously disturbs clinical decision. This study aimed to establish a solution integrated with laboratory information system to clear the bias from hemolysis on serum NSE test. Methods The reference range of serum hemolysis index (HI) was first established, and specimen hemolysis rate was compared between HI test and visual observation. NSE concentration in serum pool with normal HI was spiked with serial diluted lysates from red blood cells to deduce individual corrective equation. The agreement between individual corrective equation and original NSE test was assayed by Bland and Altman plots. Results The high HI existed in 32.6% of specimens from patients. The NSE median of hemolyzed specimens was significant higher than the baseline (p = 0.038), while the corrected NSE median had no difference compared with the baseline (p = 0.757). The mean difference of corrected NSE and initial NSE was 1.92%, the SD of difference was 5.23%, and furthermore, the difference was independent of tendency of HI (Spearman r = −0.069, p = 0.640). The 95% confidence interval of mean difference (from −8.33% to 12.17%) was less than the acceptable bias range (±20%). Conclusion The agreement between individual correction equation and NSE assay was satisfied. Our automated processing algorithm for serum NSE could provide efficient management of posttest data and correct positive bias from specimen hemolysis.


| INTRODUC TI ON
Serum neuron-specific enolase (NSE) presented exclusively in neurons and neuroendocrine tissues [1][2][3] is an important tumor marker aiding to diagnose cancer of the neuroendocrine type, in particular small cell lung cancer (SCLC) 4 and neuroblastoma (NB). 5,6 Subsequent to diagnosis, NSE is more frequently applied to evaluate the effect of therapy and to monitor metastasis or relapse of SCLC and NB. 7 In addition, serum NSE plays a paramount role in the differential diagnosis of causes of dementia, 8,9 for assessment of the severity of traumatic brain injuries, 10,11 and for prognostication of likelihood of recovery after hypoxic ischemic brain injury. 12,13 It is very important for clinical laboratory to provide a precise NSE measurement result. However, the value of serum NSE measurement is easily influenced by specimen hemolysis due to more abundant NSE in circulating red blood cells (RBCs) and platelets. 3,14 The hemolysis brings a higher false result, which disturbs clinical decision. It is well known that specimen hemolysis dominated the majority of preanalytic error (from 40% to 70%) in clinical laboratories. 15,16 In face of hemolysis influence on familiar biochemical analytes, such as potassium, some clinical laboratories prefer to reject the result and recollect specimen, 17 whereas some support to release the result with or without evaluation of the degree of hemolysis. 18 As for NSE, it is more urgent to resolve this dilemma, and because serum NSE magnitude is thousand times less than that of potassium, specimen hemolysis brought striking great influences on serum NSE. It is reported the positive interference would occur when cell-free hemoglobin concentration in serum is above 0.338 g/L at which visual inspection failed to identify. 4 Clinical researchers are looking forward to establish an effective corrective formula in order to minimize hemolysis influence from specimen. First of all, the degree of hemolysis should be tested. The measurement method of micro-cell-free hemoglobin in serum has been developed from visual inspection, 19 manual spectrophotometer 20 to automatic biochemistry analyzer whose results can be processed by the laboratory information system (LIS). 21 Visual inspection is an unreliable approach for assessing sample hemolysis, since it is arbitrary, not traceable and also plagued by poor sensitivity. 22 Recently, cell-free hemoglobin in serum is automatically quantified by absorbance measurements at different wavelengths on laboratory biochemistry analyzer, and the concentration of cell-free hemoglobin is finally reported as "hemolysis index" (HI). 23

| Measurement of serum NSE
The serum NSE was tested by electrochemiluminescence immunoassay on Cobas e 602 analyzer (Roche Diagnostics, Mannheim, Germany).
According to the manual, the 95 th percentile of reference range is 16.3μg/L, and the measurable range is from 0.05μg/L-370μg/L.

| Measurement of serum hemolysis index (HI)
Control specimen was drawn into inert separation gel vacuum collective tube by expert nurse according to the standard collection procedure. One specified laboratory technician got carefully the specimen back to the screening department laboratory, avoiding shake and jolt. The specimen was centrifuged at 1,300 ×g for 10 min at room temperature within 1 h from collection (according to the manual) and then was analyzed for serum NSE within 2 h from collection.
The HI was analyzed on Cobas c 702 biochemistry analyzer (Roche Diagnostics, Mannheim, Germany) by paired bichromatic wavelengths using Serum Index Gen2 kit (Roche Diagnostics, Mannheim, Germany).
It was calculated as the difference of absorbance at 570 nm and 600 nm converted into absolute number (range: 1-1,000). The semiquantitative value of HI corresponded to free serum hemoglobin (Hb) concentration (1 HI =10 mg/L Hb). 21 The repeatable precision and intermediate precision were 2% and 5%, respectively.

| Sensitivity comparison of hemolysis recognition between visual observation and HI test
For serum sample with HI above 95 th percentile of reference range, the hemolysis was independently classified by two technicians as the following four degrees: no hemolysis, slight hemolysis, hemolysis, and severe hemolysis. The HI test and hemolysis degree were finished in 1 day for all specimens. In addition, the severe hemolyzed serum determined by visual observation among all specimens in the laboratory was further analyzed for peak HI.

| Analyzation of NSE/Hb ratio
Concentrated RBCs lysate was prepared according to the published study. 20 It was then diluted by normal saline according to the ratio of 1:30 and centrifuged at 1,800 ×g for 10 min to remove cellular debris. The supernatant served as diluted RBCs lysate ready for measurement of NSE and HI to obtain NSE/HI ratio.

| Derivation of individual corrective equation for NSE in hemolyzed serum
Specimens with normal HI out of 503 patients were chosen from 23 serum pools whose baseline NSE concentrations ranged from low to high. Each serum pool was divided into 11 aliquots. Concentrated RBCs lysate from one healthy individual was used as spiking NSE origination, which was diluted with normal saline to serial HI concentrations: 6,000, 3,000, 1,500, 750, 375, 187.5, 93.75, 46.87, 23.43, and 11.71. These 10 serial diluted RBCs lysates were added into 10 aliquots of one serum pool according to 1:10 ratio. Normal saline was added into the eleventh aliquot serum, serving as baseline serum.

| Performance validation of individual corrective equation
We evaluated the performance of individual corrective equation on specimens from 47 inpatients. Serum of patient's specimen that had finished NSE examination served for baseline self-control. Then the specimen was intentionally hemolyzed by the following procedures: collect some (30 μl-100 μl) RBCs beneath separation gel, mix the RBCs with serum above separation gel, then aspirate 100 μl mixture of serum and RBCs to proceed two cycles of refrigeration, thaw and vibration to break the RBCs, and transfer the supernatant into the specimen. After centrifugation, the intentionally hemolyzed specimen was finished for testing serum NSE and HI.
The NSE/Hb ratio of RBCs from these inpatients was also an-

| Statistical analysis
Statistical analysis was carried out by GraphPad Prism5.0. The distribution of data set was measured by KS normality test. The serum HI and NSE concentration were described by median and interquartile range, and Mann-Whitney test was used to compare between groups for serum HI or serum NSE concentrations. Reference range was defined according to the Clinical and Laboratory Standards Institute (CLSI) document C28-A3. 24 Nonlinear regression was used to ascertain the relationship of serum HI and NSE concentration increment from RBCs. The Bland and Altman plots were used to evaluate the agreement of two measurement methods. Statistically significant was considered when a p < 0.05 for all statistical comparisons.

| Establishment of the reference range of serum NSE-specific HI
Serum NSE from the 200 cases of healthy control individuals were all below 16.3 μg/L, the median was 8.5 μg/L, and the 95 th percentile was 11.8 μg/L which was lower than 16.3 μg/L defined in the manual. The HI reference range (95%) was from 0 to 5, which was in accordance with HI operating manual, and the median of serum HI specific to NSE assay was 2.

| Abnormal HI presented in 32.6% of inpatients' serum
The serum HI values of inpatients were significantly higher than that in healthy control (p < 0.0001, Figure 1). According to HI reference range we have established, abnormal HI was defined as above 5.
Generally, abnormal HI (32.6%, 164/503) may occur in serum with various levels of NSE concentrations (Table 1), and the median of abnormal HI value was 11 (equivalent to 110 mg/L free hemoglobin) which was far less than the visual detectable hemolysis limit of 300 mg/L free hemoglobin. The abnormal serum HI rate was 21

| Corrective equation for hemolyzed serum NSE concentration must introduce personal variable: NSE/HI ratio
After spiking with RBCs lysate, 165 samples with HI >5 out of 230 serum pools contributed data pairs of NSE and HI to deduce formula. By one step nonlinear fit, a straight line equation (Figure 2) was determined for actual NSE concentration increment (y) plotted

as a function of HI meas (x):
This corresponds to 4.0 μg/L false elevated NSE originating from RBCs lysate when serum HI was 11. But it was a pity that NSE/HI ratio (R) was not introduced into this equation. The increased NSE amounts from hemolysis were theoretically different when two individuals with different NSE/HI ratios presented same serum HI value, and the difference would become larger following the increasing HI mea . Personal NSE/HI ratio (R) was already demonstrated more accurately than a mean of R from healthy individuals which was used as generalized R by some researches. 21,25 Hence a rational equation must introduce personal R as variable. When the slope and the intercept with y-axis were introduced personal R (the R of RBCs in spiking test was 0.31), the above equation is shown as: Then, by introducing individual R, HI meas and NSE meas as variables, we obtained an individualized corrective equation for serum NSE influenced by hemolysis:

| The performance of individual corrective equation accorded with quality requirements
We evaluated the equation performance of 47 specimens from validated patient group ( Table 2). The validated HI range (6-314) covered the common abnormal HI range (6-183). The validated NSE level covered the measurable range of NSE assay. The NSE median of specimens hemolyzed was significant higher than the baseline (p = 0.038), while the corrected NSE median had no statistical difference compared with the baseline (p = 0.757). The agreement between individual correction equation and NSE assay was evaluated by Bland-Altman plots (Figure 3). The mean difference of corrected NSE and initial NSE was 1.92%, the SD of difference was 5.23%, and the difference was independent of tendency of HI (Spearman r = −0.069, p = 0.640). The 95% confidence interval of mean difference was from −8.33% to 12.17%, which was less than the acceptable range of original NSE result (±20%), although the difference was slightly larger at lower NSE level than at higher level. The key of clinical application for a tumor marker focused on the level variation trend, at which the increase or decrease less than 25% had no clinical importance. These results indicated that individual correction could be an alternative option for serum NSE concentration which was influenced by specimen hemolysis.

| DISCUSS ION
Carraro first investigated the causes for specimen hemolysis in 2,000 and proved that hemolytic specimens were mostly due, up to hue. 26 Lippi also demonstrated 5.6% for hemolysis rate by visual observation in 2009. 27 However, visible inspection with poor sensitivity brings low hemolysis rate. Clinically significant false-elevations of NSE are observed even with visually undetectable hemolysis. 25 One survey employing HI test however showed an overall rate of 10.4% for hemolysis in primary healthcare centers and even a 31.1% rate in the emergency department. 28 Our data demonstrated that the visual inspection could only find minority of specimens that were hemolysis judged by HI test. Based on the serum HI reference we established, we proved 32.6% for hemolysis rate. In contrast to visual the specimen is kept before centrifugation under room temperature, the more erythrocytes are broken. In our study, the specimens were centrifuged within 1 h from collection and were tested for HI, and our reference range is high in credibility by excluding specimen hemolysis.

CO N FLI C T O F I NTE R E S T
The authors declare that they have no conflict of interest.

AUTH O R CO NTR I B UTI O N S
Jun-ye Wang and Shu-qin Dai involved in conception and design.
Shu-qin Dai contributed to administrative support. Xiao-min Liu and Xiao-hua Liu involved in provision of study materials or patients; Min-jie Mao contributed to collection and assembly of data. Yi-jun Liu involved in data analysis and interpretation. All authors involved in manuscript writing and final approval of manuscript.

DATA AVA I L A B I L I T Y S TAT E M E N T
The datasets analyzed during the current study are not publicly available due to patient privacy concerns, but are available from the corresponding author on reasonable request.