Original Article Mechanism of interference by haemolysis in the cardiac troponin T immunoassay Ravinder Sodi, Simon M Darn, Andrew S Davison, Anthony Stott, Alan Shenkin
Abstract Addresses Department of Clinical Biochemistry and Metabolic Medicine, Royal Liverpool and Broadgreen University Hospital, Prescot Street, Liverpool, L7 8XP, UK Correspondence Mr Ravinder Sodi Email:
[email protected]
Background The cardiac troponins have been shown to be sensitive and specific biochemical markers of myocardial infarction and highly prognostic for future adverse events in patients with acute coronary syndromes. There have been reports suggesting that haemolysis causes a negative interference in the cardiac troponin T (cTnT) assay but the mechanism(s) involved remain unknown. Here we show the effects of haemolysis and haemoglobin per se on the cTnT assay. Methods The effect of haemolysis was studied by the addition of prepared haemolysate to serum samples with known and clinically relevant cTnT levels. The effect of haemoglobin was studied by the addition of haemoglobin of increasing concentrations and noting its effect on the level of cTnT measured. The effect of putative proteases was determined indirectly by incubating samples with spiked cTnT with various protease inhibitors and observing the changes in the measured cTnT levels. Results The results show that both haemolysis, which is the release of haemoglobin and corpuscular contents, and haemoglobin itself negatively interfere in the cTnT assay in a concentration-dependent manner, although the former had a greater magnitude of effect. On haemolysis, indirect evidence suggests that proteases are released which degrade the cTnT in serum, thus causing the decreased levels detected. Pepstatin A, a reversible inhibitor of aspartic proteinases, effectively inhibited the loss of cTnT in serum at 371C and pH 7.4 over a 48-h period. We found that at a haemoglobin level of 0.75 g/L, cTnT declined by more than 10% of the initial concentration, suggesting that falsely decreased levels due to haemolysis may significantly affect the clinical utility of the assay. Conclusions Haemolysis, haemoglobin per se and possibly proteolysis play a role in the negative interference in cTnT assays. Measures to reduce this interference must be implemented. Ann Clin Biochem 2006; 43: 49–56
Introduction The circulating concentration of cardiac troponin T (cTnT) has been shown to be a sensitive and speci¢c biochemical marker of myocardial damage1,2 as well as a prognostic marker of future cardiac events.3 Like cardiac troponin I (cTnI), it is used for critical decision making and therefore it is imperative that all known interferents are fully characterized, and measures to eradicate them or reduce their frequency are implemented. The Roche electrochemiluminescent cTnT assay (Roche Diagnostics, UK) is the only commercially r 2006 The Association for Clinical Biochemistry
available cTnT method. It was initially believed that the assay was una¡ected by haemolysis, with the assay package insert stating a cut-o¡ of11g/L (recovery within 710% of the initial value).4 However, there have been recent reports suggesting that haemolysis causes a negative interference in this assay.5--7 The exact mechanism of this interference is not known. Previously, it had been hypothesized that during haemolysis proteases may be released from erythrocytes, which then degrade the cTnT.8 Another study has shown that haemolysis, which is the liberation of both haemoglobin and cellular contents of the cell, inter49
50
Sodi et al.
feres in the Opus immunoassay system cTnI assay (Behring Diagnostics) and not haemoglobin itself.9 Whether haemolysis, haemoglobin per se, or both interfere in the cTnT assay has not been fully established. Haemolysis is an unavoidable occurrence in the clinical laboratory.10 With the decentralization of phlebotomy services, the continued increase in the use of pneumatic tube systems,11 the drawing of blood through a needle into syringes12 and other factors, it will continue to pose a major pre-analytical problem in the foreseeable future. Thus, measures to reduce haemolysis must be implemented. To do this e¡ectively, we believe the ¢rst step is to clarify the mechanism(s) involved. Here we provide some insights into the mechanism of interference by in vitro haemolysis in the cTnT assay.
Methods Preparation of haemolysate The local institutional ethics committee approved this study as work carried out for assay development and quality control. Two tubes each of EDTA-plasma and serum-gel (Sarstedt, UK) were collected from10 normal volunteers after obtaining written consent. The blood in EDTA plasma was pooled to obtain a sizeable volume and centrifuged to collect the erythrocytes. The plasma was decanted and the separated erythrocytes were washed three times with 0.9% cold saline to remove any trace of EDTA. After each wash, the cells were collected by centrifugation for 2 min at 1000 rpm. The haemolysate was then prepared using the osmotic shock method.13 The washed erythrocytes were diluted with an equal volume of distilled water, mixed thoroughly and frozen overnight at 201C. The following day, the sample was thawed and centrifuged for 30 min to remove cellular debris. The supernatant haemolysate was decanted into a clean tube. The haemolysis index (HI), in arbitrary units (AU), of the haemolysate was measured spectrophotometrically using the Roche P-module (Roche Diagnostics, UK). The primary and secondary wavelengths used are 340 and 376 nm, respectively. Manufacturer-de¢ned factors are used to calculate the HI. In our laboratory, a cut-o¡ of 31 AU is used; results of analytes a¡ected by haemolysis are automatically cancelled if this threshold is exceeded by a sample. In this study, the haemoglobin content of the haemolysate was estimated from an in-house calibration curve where 1 HI is approximately equal to 0.024 g/L of haemoglobin and is linear to about 10 g/L. Separately, the serum-gel samples collected from volunteers and three patients were allowed to clot, pooled and centrifuged to separate the serum. An aliquot of the serum was used to con¢rm that the pool Ann Clin Biochem 2006; 43: 49–56
was cTnT negative (i.e. cTnT o0.01 mg/L [manufacturer stated cut-o¡ and 99th percentile reference limit]).14,15 We did not determine the blood group of the volunteers, as it was not feasible to do so. Furthermore, there is no evidence to suggest that the red blood cell surface antigens interfere in the cTnT assay.
To study the effect of haemolysis We studied two subjects with cTnT levels above the ROC-based traditional World Health Organization (WHO) diagnostic criteria for the diagnosis of a myocardial infarction (MI), which is 0.1 mg/L.16 We also obtained blood from a patient with a cTnT concentration close to the new diagnostic threshold with a total coef¢cient of variation (CV) of 10%, which is 0.03 mg/L.17,18 To 200 mL aliquots of these patient samples, 20 mL of the haemolysate was added after serial dilution with cTnT negative serum to obtain a ¢nal HI, after correcting for the dilution, ranging from a control with no haemolysate added to approximately 200 AU. cTnT was measured on the Roche E170 (Roche Diagnostics, UK). The internal quality control and external quality assurance were within acceptable limits (o2 standard deviations of the accepted mean) throughout the study. All analyses were carried out in triplicate, in one batch overnight, to minimize inter-assay variability.
To study the effect of haemoglobin To test whether haemoglobin itself was responsible for the interference, we prepared puri¢ed human haemoglobin (Sigma, UK) by solubilizing it in cTnT negative serum to which recombinant human cTnT (Sigma, UK) at a concentration of 12 and 0.1 mg/L had been added. The rationale for these concentrations are discussed later. The ¢nal haemoglobin concentrations ranged from a control with no added haemoglobin to 10 g/L. Again, the cTnT levels were measured in triplicated in one batch on the Roche E170 (Roche Diagnostics, UK).
To study the effect of proteolysis To test the hypothesis that, on haemolysis, proteases may be released from the erythrocytes that degrade the cTnT in serum, we examined the e¡ect of various protease inhibitors on cTnT degradation in vitro. This was driven by reports of the presence of proteases such as cathepsin E in human erythrocytes.19 We investigated the e¡ect of pefabloc (irreversible inhibitor of serine proteases), antipain hydrochloride (reversible inhibitor of serine and cysteine proteases including some trypsin-like serine proteases) and pepstatin A (reversible inhibitor of aspartic proteases, cathepsins, pepsin and renin). All of these inhibitors were purchased from Sigma-Aldrich (UK) and the working
Interference by haemolysis in the cTnT assay
Audit of the incidence of haemolysis Data for all the haemolysed samples over a period of one year (June 2004 -- June 2005) was retrieved from the local information database (Telepath). From this data, we calculated the percentage of requests for cTnT that were haemolysed. Based on the above studies, some predictions were made.
Analysis of cardiac troponin T In this study, cTnT was determined using the thirdgeneration electrochemiluminescent immunoassay (ECLIA) on the Roche E170 immunoassay analyzer (Roche Diagnostics, UK). The assay utilizes the ‘sandwich’ principle, with a biotinylated monoclonal cTnTspeci¢c ‘capture’ antibody and a ‘detection’ antibody labelled with a ruthenium complex, which is also cTnT-speci¢c. The third-generation assay has been standardized using recombinant human cTnT to improve assay linearity. The total duration of the assay is 18 min. It has a measuring range of 0.010--25.00 mg/L, de¢ned as the lower detection limit of the assay and the maximum of the master curve.14 In the latest version of the assay, the manufacturers quote 0.03 mg/L as the
concentration at which a total imprecision (CV) of 10% is achievable.14
Statistical analyses In all cases, means and standard errors (SEM) of triplicates are shown unless otherwise stated in ¢gure legends. To examine the e¡ect of haemoglobin on the cTnT assay, the Kruskall-Wallis one-way analysis of variance on ranks test was used followed by Dunn’s method (SigmaStat, Systat software, UK). Statistical signi¢cance was considered if Po0.05. Recovery in the proteolysis studies was calculated with respect to the measured cTnT level in unhaemolysed control samples at a given time point, that is, unhaemolysed control samples at each time point were considered 100%.
Results Effect of haemolysis The e¡ect of haemolysis on the cTnT assay is shown in Figure 1. Two subjects with cTnT levels above the diagnostic threshold for an MI of 0.1 mg/L16 and one subject with a cTnT level of 0.03 mg/L, which is the level with a 10% total CV, were studied.17,18 In all cases, the cTnT concentration decreased with increasing HI (and therefore haemoglobin levels). At a HI index of 50, there was almost a 50% decrease. Our data is in agreement with previously published reports using a
0.5 Troponin T concentration (µg / L)
concentrations used were as per the manufacturer’s instruction (1, 14 and 1 mM, respectively). They were all dissolved in serum except pepstatin A, which as per the manufacturer’s instructions was dissolved in ethanol 90%:10% acetic acid. A control experiment with ethanol 90%:10% acetic acid was therefore also undertaken. Two sets of experiments were carried out simultaneously. In one experiment, haemolysate was added to serum to a ¢nal HI of approximately 600 AU (HS). We chose this high level of haemolysis based on the ¢ndings of the initial experiments detailed above. In the second experiment, plain serum with no added haemolysate (UHS) was used. Both the HS and UHS were spiked with recombinant human cTnT (Sigma-Aldrich, UK) at a concentration of 10 mg/L. The rationale for this level is discussed below. This pool was further subdivided into ¢ve; one set each to which pefabloc, antipain HCL and pepstatin A were added at a ¢nal volume of 5 mL. In one experiment, 5 mL ethanol 90%:10% acetic was added and one set was used as a control experiment to which 5 mL of 0.9% saline had been added. In all cases, the inhibitors were added before the addition of cTnT. Each experiment was also carried out at three di¡erent temperatures: 371C, 41C and at room temperature (RT, 20--251C). After a baseline pre-treatment cTnT measurement, aliquots from each set were taken every 12 h for the ¢rst 24 h and then once every day for a period of 6 days (144 h). The aliquots were frozen at 201C until analysis.
51
Patient 1 Patient 2 Patient 3
0.4
0.3
0.2
0.1
0.0
0
50 100 150 200 Haemolysis Index (arbitrary units)
250
Figure 1 Effect of haemolysis on the cardiac troponin T levels in serum. Haemolysis index (HI) was measured spectrophotometrically on the Roche p-module. Three patient examples are shown with different initial cardiac troponin T (cTnT) concentrations. The horizontal line is the diagnostic threshold based on the new definition of myocardial infarction17,18 and the dashed line is the threshold based on the traditional ROC-derived cut-off.16 At each HI, cTnT was measured in triplicate and the mean and standard error of the mean are shown Ann Clin Biochem 2006; 43: 49–56
Sodi et al. Potassium Haptoglobin Cardiac troponin T
300
% change
200
100
0
−100
0
2
4 6 8 10 Haemoglobin concentration (g/ L)
12
Figure 2 Effect of haemolysis on potassium, haptoglobin and cardiac troponin T (cTnT). The initial cTnT level was 0.1 mg/L. Mean of duplicates are shown di¡erent strategy to prepare the haemolysate, and our haemoglobin range encompasses those reported in the previous studies.5,6 Thus, we concluded that there is a negative bias in the measurement of cTnT with increasing levels of haemolysis in serum. The investigation of a haemolysis e¡ect on CK-MB on the same Roche E170 platform (Roche Diagnostics, UK) revealed no interference. The current CK-MB product insert states a haemolysis threshold of 15 g/L (Hb ¼ 0.932 mmol/L), above which haemolysis has an e¡ect.20 Thus, any observable e¡ect is attributable to e¡ects other than on the electrochemiluminescent detection system used in the Roche E170 instrumentation. In one experiment shown in Figure 2, we measured potassium to serve as a positive control to demonstrate in vitro haemolysis. As can be seen, with increasing haemolysis (as ascertained by the haemoglobin level), the potassium levels rose from 3.5 to 10.5 mmol/L, an increase of 300%. At the same time, the cTnT levels decreased by more than 50% from a starting point of 0.1 mg/L. To demonstrate that the haemolysis was strictly due to in vitro causes, we measured haptoglobin. In vivo causes of haemolysis would have caused a decrease in the haptoglobin levels,21 which is not the case here. Taken together, the observed decrease in the cTnT levels is due to an in vitro cause of haemolysis in view of the fact that the potassium concentrations were increased and haptoglobin levels remained constant. To quantify the magnitude of the e¡ect of the interference by haemolysis in the cTnT assay, we constructed an interferogram as previously described.22 Figure 3 shows that with increasing haemolysis index (and haemoglobin levels), there was a proportionate decrease in the cTnT levels. A decrease of more than 10% from an initial value of 0.1 mg/L in the cTnT levels was Ann Clin Biochem 2006; 43: 49–56
Mean % change in cTnT concentration
52
0
0.78
1.56
3.13
6.25
12.5
0
37.5
76.5
154
316
632
0 −5 −10 −15 −20 −25 −30 −35 −40
Haemoglobin levels (g / L); haemolysis index (AU)
Figure 3 Interferogram showing the magnitude of the effect of haemolysis on the cardiac troponin T assay. A decrease of more than 10% from an initial level of 0.1 mg/L was seen at the haemolysis index of approximately 31 arbitrary units (AU), which is the haemolysis threshold in our clinical laboratory using the Roche Modular Analytics unit. Note that in order to show both the haemolysis index (bottom) and the corresponding haemoglobin levels (top) on the same graph the x-axis is categorical and non-linear
seen at an HI of approximately 31, which is the haemolysis threshold used on the Roche Modular Analytics unit (Roche Diagnostics, UK) in our laboratory. This suggests that cTnT should not be measured in any samples exceeding the haemolysis threshold because it can potentially lead to an increased incidence of false negative results. A HI of 31 is approximately equivalent to a haemoglobin level of 0.75 g/L. Incidentally, this was close to the recommended 1.0 g/L (0.062 mmol/L) cut-o¡ in the Roche product alert informing users that haemolysis interferes in the assay and that cTnT should not be measured in haemolysed samples with haemoglobin greater than this level.23
Effect of haemoglobin As shown in Figure 4a and b, with increasing concentration of haemoglobin, there was a statistically signi¢cant decrease in the measured cTnT level compared to the initial concentration of 12 mg/L (Po0.001) and 0.1 mg/L (P ¼ 0.027). Based on this ¢nding, we concluded that both haemolysis and haemoglobin had a negative e¡ect on the Roche cTnT assay. This ¢nding is consistent with a previous study, which concluded that haemoglobin released during haemolysis mediates the interference in the cTnT assay using the Elecsys 2010 Immunoanalyser.6 We noted that there was a greater degree of decline in cTnT levels in the haemolysis experiments using haemolysate (Figure 1) compared to experiments in which
Interference by haemolysis in the cTnT assay 120
(a) 0.12
100 Mean % recovery
Troponin T concentration (µg /L)
53
0.10 * * 0.08
80 60 40 20
0.06
0 0
10
20
30 40 Time (hours)
50
60
0.00 0
2 4 6 8 10 Haemoglobin concentration (g / L)
12
UHS control
UHS + eth /AA
HS + pepA
HS + eth /AA
HS control
Troponin T concentration (µg / L)
(b) 12
Figure 5 Effect of pepstatin A on cardiac troponin T at 371C. Mean percentage recovery of a representative experiment is shown. HS – haemolysed samples, UHS – unhaemolysed samples. PepA – pepstatin A (1 mM), eth/AA – ethanol/acetic acid
11 *
10
* 9
*
* *
8 0 0
2 4 6 8 10 Haemoglobin concentration (g / L)
12
Figure 4 Effect of haemoglobin on the cardiac troponin T (cTnT) levels in serum. (a) Troponin T spiked at 0.1 mg/L (P ¼ 0.027 by Kruskall-Wallis test; multiple comparisons versus initial level by Dunn’s method where denotes P values were o0.05). (b) Troponin T spiked at 12 mg/L (Po0.001 by Kruskall-Wallis test; multiple comparisons versus initial level by Dunn’s method where denotes P values were o0.05). At each concentration of haemoglobin, cTnT was measured in triplicate and the mean and standard error of the mean are shown
haemoglobin was added (Figure 4). Thus, for a given HI, for example 50 AU (equivalent to 1.2 g/L of haemoglobin), cTnT fell by almost 50 % from the initial concentration in the haemolysis experiment (using haemolysate) as shown in Figure 1 (patients 1 and 2) but only by 10--16% in Figures 3 and 4a.
Effect of proteolysis No observable e¡ect was seen with pefabloc or antipain HCL (data not shown). No e¡ect was observed at RT or at 41C for any of the inhibitors (data not shown). As shown in Figure 5, there was a decrease of between 30--50% in cTnT in samples with added haemolysate (HS, circle) compared to the respective unhaemolysed
samples (UHS, diamond). The 90% ethanol: 10% acetic acid did not have any signi¢cant e¡ect on cTnT in either the HS (triangle) or UHS (square) samples at all three temperatures tested. In the UHS control samples (diamonds), no loss in cTnT was observable over the entire 6 days (144 h) compared to the initial level. For the HS samples, we were unable to con¢dently quantify any e¡ect after 48 h because the measured cTnT at 371C started to decline, possibly due to sample degradation and cTnT instability. However, we found that at 371C, pepstatin A reduced the loss of measured cTnT due to proteolysis (upon haemolysis) over the entire 48-h time period (Figure 5, crosses).
Audit of the incidence of haemolysis In an audit of cTnT requests over a period of 1 year, we found that 12,287 requests had been made for the test. From the total requests, 781 were haemolysed for various reasons, giving an incidence of haemolysis of 6.4% in this particular population. It was di⁄cult to retrospectively assess whether repeats were received because in many cases patients had been transferred to a di¡erent ward and the practice at our hospital is such that di¡erent locations might use a unique number even for the same patient. In addition, it was noted that in many cases, repeats were not received soon enough to allow us to take into account the kinetics of cTnT release. This delay would have invalidated any comparative analyses undertaken. It was therefore impossible to meaningfully compare the concentration of cTnT in haemolysed samples (which we measured but did not report) and the concentration Ann Clin Biochem 2006; 43: 49–56
54
Sodi et al.
of cTnT in the repeat unhaemolysed sample. Therefore, it is impossible to ascertain the magnitude of the e¡ect of haemolysis in actual practice in the current study. However, based on the ¢ndings here in patients at clinically appropriate levels (diagnostic threshold of 0.1 mg/ L16 and level with a total CV of 10%, which is 0.03 mg/ L17,18) it is not unreasonable to predict that an increased incidence of false negatives could occur. In the two patients with initial cTnT of 0.226 and 0.037 mg/L shown in Figure 1, at a HI of about 50 and 5, respectively, the diagnosis would have changed from MI (assuming the presence of clinical symptoms and/or characteristic electrocardiographic changes) to highrisk cTnT level of prognostic signi¢cance and within 99% percentile reference limit respectively. This is discussed further below.
Discussion Our results suggest that the interferent is released on haemolysis and is e¡ectively inhibited by pepstatin A at 371C. Although a number of proteases can be inhibited by pepstatin A, we are not aware of any proteases present in erythrocytes other than cathepsin E.13 That pepstatin A works only at 371C makes this ¢nding of little immediate practical application. However, it raises important questions regarding the clearance of troponin T in vivo where the temperature is 371C. It has been shown that the aspartic proteinase which include the cathepsins have an optimal working pH of about 4.5--5.5.24 Here we did not investigate the e¡ects of pH because this does not have a bearing in a realistic situation. At a physiological pH of approximately 7.4, which was maintained throughout this study, we observed that only pepstatin A exerted any discernible effect. The modulatory action of pH and temperature is a recurrent theme in enzymology; however, we are not aware of any reports showing the temperature dependence of either the activity of cathepsin E or its speci¢c optimal working pH. Thus, this demonstration here might be an important ¢nding in itself and merits further investigation. The susceptibility of cTnI to proteolysis is well documented.25 Free cTnI is rapidly degraded by proteases present in serum, especially after necrosis. The study by Katrukha et al.25 showed that less than 10% of the intact cTnI remained in serum after 20 h of incubation at 371C. Unlike cTnI, there is no data available to our knowledge demonstrating the degradation of cTnT in this context. However, it has been shown that cTnT is more stable in serum, and reports by Collinson et al.26 indicate that it is stable at room temperature, at 41C, up to 5-years at 701C and even after ¢ve freeze-thaw cycles. Here we report that in haemolysed samples, cTnT is probably rapidly degraded and the ¢nding that Ann Clin Biochem 2006; 43: 49–56
pepstatin A can prevent this loss at 371C suggests, in part, a proteolytic mechanism. Haemoglobin clearly also caused the decrease in cTnT in a concentration-dependent manner. We note that the interactions between haemoglobin and haptoglobin can provide vital clues regarding the binding of troponin T to haemoglobin. Is it possible that pepstatin A can disrupt this interaction? Can haptoglobin be included in the assay reagents to ‘mop up’ the released haemoglobin? Further work is required to answer these questions. It is di⁄cult to estimate how much of a contribution comes from haemoglobin itself, the haemolytic process or proteolysis. However, based on the di¡erent rates of change observable for haemolysis and haemoglobin per se, it seems apparent that the e¡ect is additive in the former where there is release of both haemoglobin and other cellular contents. It is not surprising that pepstatin A did not fully prevent the cTnT degradation. As we have already stated, proteolysis is only part of the mechanism causing this degradation. Other factors are clearly involved, which pepstatin A has no e¡ect on. We stress that there is a need to look for solutions to this problem of haemolysis interference, especially for those analytes required for critical-decision-making such as cTnT used for the diagnosis of MI in the appropriate setting and for risk strati¢cation.27 At our hospital, we have shown that 6.4% of all cTnT requests in one year were haemolysed for one reason or another. Therefore, the rejection of haemolysed samples for cTnT may not be a viable option in the long run. Possible solutions to this problem might be the use of sample tubes with‘cocktails’of inhibitors to prevent proteolysis as is the case for B-type natriuretic peptide28 or the incorporation of the inhibitor(s) in the assay reagents. Another solution suggested by a recent study was the mathematical correction of the biased cTnT results using HI.6 These former solutions might result in the reduction of the degradation of cTnT due to haemolysis and will decrease the need to obtain repeat samples from patients, who in most cases will be in a critical condition. On the other hand, we concede that the requirement for ‘cocktail’ tubes, if found to be e¡ective, will increase the cost of providing this important test. It might also be argued that it may be better to request a repeat and obtain a result with full con¢dence rather than mathematically correct for the haemolysisinduced decrease. Regarding our experimental approach, we added recombinant cTnT at very high concentrations in some experiments to delineate the e¡ects of haemoglobin and the protease inhibitors as previously undertaken in a similar context.29 We took into account the limit of detection of the assay (0.01 mg/L), the level at which the total precision of 10% or less (coe⁄cient of variation) was achievable, that is, 0.04 mg/L30 or 0.03 mg/L
Interference by haemolysis in the cTnT assay
as stated by the manufacturer14 and the need to magnify any e¡ect if present. Importantly, we also repeated the same using the traditional clinically diagnostic level of 0.1 mg/L.16 In addition, other studies corroborate our ¢ndings,5--7 and using actual patient samples with varying cTnT levels (Figure 1) we were able to document an e¡ect by haemolysis. It has been suggested that the use of haemolysate prepared using the osmotic shock method does not mimick the occurrence of haemolysis in a real clinical situation.31 A recent study used a syringe to induce haemolysis in a sample.31 We stress that our methodology is the widely accepted mode of demonstrating an e¡ect by haemolysis in assays.5--7,13,22 Furthermore, we argue that while the syringe method is the ‘physically’ realistic way of inducing haemolysis as it mimicks an operators mechanical action, it is not ‘physiologically’ relevant because cells are rarely subjected to sheer mechanical stress and, in most cases, it does not ensure the lysis of all cells, which may obscure the e¡ect of haemolysis, if any. In a pilot study, we checked for the presence of interference by documenting non-linearity on dilution. Importantly, as we stated in the methods section, we corrected for the small dilution factor, which we found was not signi¢cant anyway. What is the potential impact of the haemolysis-induced decrease of cTnT in actual clinical practice? As stated in the results section, we could not retrospectively compare results obtained in a haemolysed sample to the unhaemolysed repeat, so as to calculate the associated clinical sensitivity. According to the new de¢nition, MI is diagnosed when blood levels of cardiac troponins are above the 99th percentile of the reference limit of an healthy population in the clinical setting of acute ischaemia.17 For cTnT, the 99th percentile reference limit was found to be o0.01 mg/L.14,15 However, it has been recommended that the cardiac troponin concentration that meets 10% CV should be used as a diagnostic cut-o¡ until more sensitive assays become available. For cTnT, this level is 0.03 mg/L.18 In this study, we evaluated the e¡ect of haemolysis in a patient with a measured cTnT level of approximately 0.03 mg/L (Figure 1) in concordance with the new de¢nition of MI.We also studied patients with higher levels of 0.226 and 0.409 mg/L (Figure 1) in keeping with the traditional WHO criteria for the de¢nition of acute MI, with a clinical discriminator value for cTnT of 0.1 mg/L according to ROC analysis.16 We also used the 0.1 mg/L level to study the e¡ect of haemoglobin on cTnT (Figure 4a). As would be expected, at concentrations near the diagnostic thresholds as shown in Figure 1 and 4a, the potential for misclassi¢cation based on both criteria was increased, at even lower haemolysis and/or haemoglobin levels. For example, in the case of patient 2 (Figure 1) with a cTnT level of 0.226 mg/L, documented ST-segment elevation on ECG was consistent with
55
a working diagnosis of MI. In a real scenario, had his HI exceeded approximately 50 AU, the measured cTnT concentration would have been less than 0.1 mg/L. Theoretically, this could have resulted in a misclassi¢cation according to the traditional criteria.16 Similarly, had the HI been greater than 200 AU, which is visible haemolysis and the sample probably would not have been accepted for analysis, the cTnT measured would have been less than the 0.03 mg/L cut-o¡, potentially ruling out MI. The same argument applies for patient 3, but using the new de¢nition of MI.17,18 A previous study by Lyon et al.6 showed that with every 1g/L increase in haemoglobin, the probability of a cTnT result 40.1 mg/L was decreased by 2.5%. Thus if a sample contains 10 g/L of haemoglobin, the probability of a cTnT 40.1 mg/L will be decreased by 25%. Another study by Snyder et al.7 demonstrated that the negative interference of added haemolysate on cTnT was clinically signi¢cant, that is greater than 30%, at a haemolysis index of 5.14 g/L. This is similar to our ¢nding, where at a similar HI we documented decrease of approximately 25% (Figure 3). As described in the results section, we noted a di¡erent magnitude of decrease in cTnT induced by haemolysis and haemoglobin per se possibly due to the release of other factors including proteases acting in an additive manner in the former. Taken together, this study and the previous ones have demonstrated that haemolysis above the accepted threshold causes a signi¢cant negative bias in the level of cTnT measured. This has the potential of causing a misclassi¢cation. One limitation of this study was that only a few patients were studied due to pragmatic and logistical considerations. Future studies using a larger sample size with a prospective design should seek to con¢rm the ¢ndings suggested here regarding misclassi¢cation. We emphasise that the e¡ects shown here are at best indirect insights, and further studies demonstrating that proteases degrade cTnTcausing a decrease in their immunoreactivity need to complement the ¢ndings suggested here. Furthermore, studies showing that haemoglobin can bind to cTnT or mask the epitopes on cTnT preventing the assay antibodies from recognizing cTnT in serum also need to be undertaken. It has been demonstrated that cTnT is released as a complex with cTnI and troponin C (cTnC) as well as free cTnT.29 Whether cTnT in the complex, free cTnT or both are a¡ected remains to be shown. However, it must be borne in mind that the current Roche cTnT assay measures free cTnT as well as the binary and ternary complexes.14 Also, contingent on ethical approval, the e¡ect of pepstatin A on cTnT in actual patient samples needs to be shown. In conclusion, here we have provided further evidence that both haemolysis and haemoglobin per se cause a decrease of cTnT in serum samples. We have Ann Clin Biochem 2006; 43: 49–56
56
Sodi et al.
shown that there is a di¡erential magnitude in the e¡ect of haemolysis and haemoglobin, with the former causing a larger decrease in measured cTnT compared to the latter. We have shown that during haemolysis, proteases such as cathepsin E may be released that degrade the cTnT. This degradation is e¡ectively reduced by the inhibitor pepstatin A at 371C for over a 48-h period. This may form the basis of a potential solution for the problem of haemolysis in the cardiac troponin T immunoassay.
Acknowledgements We thank Mr John Dutton and Mrs Sue Levine for carrying out the cTnT assays in this study.We acknowledge Mr Trevor Hine for extracting the audit data used in this study.We thank Roche Diagnostics Incorporaton for providing the kits for the cTnT assay.
References 1 Katus HA, Remppis A, Neumann FJ, et al. Diagnostic efficiency of troponin T measurements in acute myocardial infarction. Circulation 1991; 83: 902–12 2 Gerhardt W, Ljungdahl L. Troponin T: a sensitive and specific diagnostic and prognostic marker of myocardial damage. Clin Chim Acta 1998; 272: 47–57 3 Hamm CW, Ravkilde J, Gerhardt W, et al. The prognostic value of serum troponin T in unstable angina. N Engl J Med 1992; 327: 146–50 4 Roche Diagnostics Troponin T STAT kit [package insert]. Elecsys Systems 1010/2010. 2002 5 Hammett-Stabler CA, Synder JA, Chapman JF, Rogers MW, King MS, Phillips JC. Hemolysis interferes with troponin I and troponin T immunoassays [Abstract]. Clin Chem 2003; 49: A89 6 Lyon ME, Ball CL, Krause RD, Slotsve GA, Lyon AW. Effect of hemolysis on cardiac troponin T determination by the Elecsys 2010 immunoanalyzer. Clin Biochem 2004; 37: 698–701 7 Snyder JA, Rogers MW, King MS, Phillips JC, Chapman JF, Hammett-Stabler CA. The impact of hemolysis on Ortho-Clinical Diagnostic’s Eci and Roche’s Elecsys immunoassay systems. Clin Chim Acta 2004; 348: 181–7 8 Sodi R, Darn S, Stott A. Time for troponin T? Implications from newly elucidated structure. Clin Chem 2004; 50: 786–7 9 Wenk R. Mechanism of interference by hemolysis in immunoassays and requirements for sample quality. Clin Chem 1998; 44: 2554 10 Jones BA, Calam RR, Howanitz PJ. Chemistry specimen acceptability. A college of American Pathologists Q-probes study of 453 laboratories. Arch Pathol Lab Med 1997; 121: 19–26 11 Sodi R, Darn SM, Stott A. Pneumatic tube system induced hemolysis: assessing sample type susceptibility to haemolysis. Ann Clin Biochem 2004; 41: 237–40 12 Carraro P, Servidio G, Plebani M. Hemolyzed specimens: a reason for rejection or a clinical challenge? Clin Chem 2000; 46: 306–7 13 Interference Testing in Clinical Chemistry. NCCLS Document EP7-P6 (13). 1986
Ann Clin Biochem 2006; 43: 49–56
14 Roche Diagnostics Troponin T Kit [package insert]. Roche Elecsys 1010/2010/ MODULAR ANALYTICS E170 (Elecsys module) immunoanalyzers. 2004, (August) 15 Apple FS, Quist HE, Doyle PJ, Otto AP, Murakami MM. Plasma 99th percentile reference limits for cardiac troponin and creatine kinase MB mass for use with European Society of Cardiology/ American College of Cardiology consensus recommendations. Clin Chem 2003; 49: 1331–6 16 Muller-Bardorff M, Hallermayer K, Schroder A, et al. Improved troponin T ELISA specific for cardiac troponin T isoform: assay development and analytical and clinical validation. Clin Chem 1997; 43: 458–66 17 The Joint European Society of Cardiology/American College of Cardiology Committee. Myocardial infarction redefined – a consensus document of the Joint European Society of Cardiology/ American College of Cardiology Committee for the redefinition of myocardial infarction. J Am Coll Cardiol 2000; 36: 959–69 18 Apple FS, Wu AHB. Myocardial infarction redefined: role of cardiac troponin testing. Clin Chem 2001; 47: 377–9 19 Yonezawa S, Nakamura K. Species-specific distribution of cathepsin E in mammalian blood cells. Biochim Biophys Acta 1991; 1073: 155–60 20 Roche Diagnostics CK-MB kit [package insert]. Roche Elecsys 1010/2010/ MODULAR ANALYTICS E170 (Elecsys module) immunoanalyzers. 2004, (August) 21 Guder WG, da Fonseca-Wollheim F, Heil W, et al. The haemolytic, icteric and lipemic sample recommendations regarding their recognition and prevention of clinically relevant interferences. J Lab Med 2000; 24: 357–64 22 Glick MR, Ryder KW, Jackson SA. Graphical comparisons of interferences in clinical chemistry instrumentation. Clin Chem 1986; 32: 470–5 23 Roche Diagnostics Technical Update. Notice of hemolysis effect on troponin T. 2003, (May) 24 Cooper JB. Aspartic proteinases in disease: a structural perspective. Curr Drug Targets 2002; 3: 155–73 25 Katrukha AG, Bereznikova AV, Filatov VL, et al. Degradation of cardiac troponin I: implications for reliable immunodetection. Clin Chem 1998; 44: 2433–40 26 Collinson PO, Boa FG, Gaze DC. Measurement of cardiac troponin. Ann Clin Biochem 2001; 38: 423–49 27 Ohman EM, Armstrong PW, Christenson RH, et al. Risk stratification with admission cardiac troponin T levels in acute myocardial ischemia. N Engl J Med 1996; 335: 1333–41 28 Belenky A, Smith A, Zhang B, et al. The effect of class-specific protease inhibitors on the stabilization of B-type natriuretic peptide in human plasma. Clin Chim Acta 2004; 340: 163–72 29 Wu AHB, Feng Y-J, Moore R, et al. Characterization of cardiac troponin subunit release into serum after acute myocardial infarction and comparison of assays for troponin T and I. Clin Chem 1998; 44: 1198–208 30 Panteghini M, Pagani F, Yeo K-TJ, et al. Evaluation of imprecision for cardiac troponin assays at low-range concentration. Clin Chem 2004; 50: 327–32 31 Dimeski G, Clague AE, Hickman PE. Correction and reporting of potassium results in haemolysed samples. Ann Clin Biochem 2005; 42: 119–23
Accepted for publication 28 October 2005
