Original Article Decision limit for troponin I and assay performance Abstract Address Background Correspondence Methods Results Conclusions

Size: px

Start display at page:

Download "Original Article Decision limit for troponin I and assay performance Abstract Address Background Correspondence Methods Results Conclusions"

Marvin Wade
5 years ago
Views:

1 Decision limit for troponin I and assay performance Paul Sheehan, John Blennerhassett and Samuel D Vasikaran Original Article Abstract Address Core Clinical Pathology and Biochemistry Division of Laboratory Medicine Royal Perth Hospital Wellington Street Perth 6000 Australia Correspondence Mr Paul Sheehan paulsheehan@health.wa.gov.au Background The rede nition of acute myocardial infarction by the Joint European Society of Cardiology and American College of Cardiology places troponin at the centre of the diagnostic strategy, in addition to lowering the diagnostic cut-off to the 99th centile of a healthy reference population. The required percentage coef cient of variation (%CV) for the assay at this level is 10. Recent publications have examined the utility of the Bayer ACS:180 troponin I assay at a cut-off of 0 1 mg/l to risk-stratify patients with non-st-elevation myocardial infarction. Methods This study examines the appropriateness of using this assay at this cutoff in individual patients. It also examines the functional sensitivity of the assay and assesses the impact of sample quality on assay performance. Results At the decision limit of 0 1 mg/l, 8% of patients would be assigned to a different risk group on repeat analysis of the same sample on the ACS:180 due to assay imprecision. The functional sensitivity (at inter-assay %CV ˆ 10) of the ACS:180 troponin I assay was determined to be 0 27 mg/l. Nineteen of 4850 routine samples (0 39%) failed duplicate precision checks as a result of poor sample quality; this was usually due to the presence of small brin particles. Conclusions Careful attention to sample quality is vital in troponin I measurement. The use of the Bayer assay for risk strati cation at a cut-off of 0 1 mg/l can lead to inconsistency of risk assessment in a small but signi cant proportion of cases. Introduction Cardiac troponins are the best biochemical markers for the diagnosis, risk strati cation and selection of therapy in acute coronary syndromes. 1^4 The paramount importance of these markers has been recognized by the National Academy of Clinical Biochemistry, the American Heart Association and the Joint European Society of Cardiology/American College of Cardiology (ESC/ACC) committee in their recommendations and guidelines. 5^7 The measurement of troponin provides diagnostic and prognostic information that is not available by other means. As the clinical role of troponin continues to evolve, there is understandably a greater focus on the ability of current troponin assays to support this role. The recent ESC/ACC consensus document on the rede nition of myocardial infarction 7 (MI) recommended the use of a single troponin concentration cut-o. This was de ned as a measurement exceeding the 99th centile of a reference control group. Acceptable imprecision for troponin assays at this cut-o was de ned by the ESC/ACC committee as 510%. Although standardization of cardiac troponin I (ctni) assays is deservedly a recognized priority, we believe that insu cient attention has been paid to the need to improve the precision of ctni assays in order to use troponin measurements e ectively in risk strati cation at low levels. This rede nition based on troponin concentration should serve to focus the e orts of manufacturers and laboratory sta on ways to improve the performance of methods for this key analyte. The FDA has recently given approval for the use of the Bayer Diagnostics ACS:180, ADVIA Centaur and Immuno 1 2 ctni assays for risk strati cation of patients with chest pain (customer bulletin no. TN ). Recent publications have assessed the potential of these and other commercial assays to riskstratify patients with non-st-elevation acute 2002 The Association of Clinical Biochemists 231

2 232 Sheehan et al. coronary syndromes at a decision limit of 0 1 mg/l. 3,8 We present data from our laboratory describing the reproducibility of the Bayer Diagnostics ASC:180 ctni assay and the impact of using di erent lower decision limits on the consistency of risk strati cation. Materials and methods All troponin measurements were performed on blood samples collected from patients into Vacutainer PST 2 gel tubes (Beckton Dickinson, Plymouth, UK). ctni was measured on the ACS:180 analyser. The minimum detectable concentration (MDC) of the assay (de ned as the concentration giving an analytical signal 2 standard deviations greater than the mean of 20 replicates of the zero calibrator) is reported by Bayer to be 0 07 mg/l. No information is provided in the kit insert on the functional sensitivity of the assay. The ACS:180 reports ctni results in the range 0^50 mg/l for undiluted samples. Study 1 Eleven samples with ctni concentrations in the range 0 0^0 8 mg/l were each analysed ten times for ctni on the ASC:180 (Bayer Diagnostics, Tarrytown NY, USA) to determine the functional sensitivity of the assay. Functional sensitivity was de ned as the lowest level at which the inter-assay imprecision, as percentage coe cient of variation (%CV), was 10. Study 2 In April 2000, Bayer Diagnostics issued a product alert to its customers following reports of falsely elevated values from some users of their ctni assay. Between 7 April and 28 July 2000, while Bayer were resolving this issue, all ctni measurements performed on an ASC:180 analyser for routine requests on patients managed at our institution were analysed in duplicate (4850 samples, 9700 assays). The precision of the duplicate pp measurements was calculated using the equation d 2 /2n (where d is the di erence between the duplicates for each sample and n is the number of samples). In addition, the data were grouped into bins to obtain the precision pro le of the assay. The impact of analytical imprecision on the consistency of risk assessment was also assessed at di erent cut-o s by examining how often one of the duplicates was on the opposite side of the selected cut-o to its partner. The e ect of pre-analytical and sample-related variables on the performance of the assay was examined by comparing sample duplicate imprecision to prede ned imprecision limits (%CV 420, absolute delta mg/l) and re-assaying all the samples that failed this test. Samples failing this test were also inspected for the presence of haemolysis, brin CV% ctnl (mg/l) Figure 1. Functional sensitivity of the Bayer ACS:180 assay for cardiac troponin I (ctni), showing inter-assay precision of ctni with patient samples (n ˆ 10). %CV ˆ percentage coef cient of variation. particles, etc., to assess the impact of sample-related variables on result quality. In a separate experiment the e ect of brin was further investigated by adding a suspension of brin particles to a patient s sample. The sample was then divided into two aliquots, one of which was centrifuged prior to analysis and one of which was mixed prior to troponin measurements. The ASC:180 ctni assay continued to perform to the manufacturer s speci cations, and internal qualitycontrol and external quality assurance performances were satisfactory throughout this study period. Results Study 1 The precision pro le of the Bayer ctni assay obtained by replicate analysis of patient samples (n ˆ 10 assays) is shown in Fig. 1. The functional sensitivity (at interassay %CV ˆ 10) was de ned as 0 27 mg/l based on these data. Study 2 A total of 4850 routine patient samples were assayed in duplicate during this study. There were three calibrations, and two reagent and two calibrator lot number changes during the period of this study. Figure 2 demonstrates the precision pro le of the Bayer ASC:180 ctni assay in the clinically important range 0 00^2 00 mg/l for duplicate samples. The precision pro le obtained by grouping the patient data into bins and calculating the %CV is illustrated in Fig. 3. This represents a robust estimate of the stable long-term performance of the assay under routine conditions. Table 1 examines the impact of assay imprecision on risk assessment and indicates the percentage of patients who would be assigned to a di erent risk

3 Decision limit for troponin I and assay performance 233 Figure 2. Precision pro le of the Bayer ACS:180 assay for cardiac troponin I (ctni) ( mg/l; n ˆ 2993). Data for samples that failed the replicate precision check due to the presence of brin clots or cell debris were deleted (n ˆ 14). group on repeat analysis of the same sample, due to analytical imprecision at di erent troponin concentration cut-o s. The estimates shown represent a bestcase scenario since intra-assay variability only is assessed. Day-to-day variation in prediction is likely to be higher in practice when the impact of inter-assay and instrument-related factors is considered.. Because of the reported problems with falsepositive ctni results during the study period, the duplicate results were compared against predetermined precision limits. Those samples failing this test were re-assayed and re-inspected. Overall, 19 (0 39%) samples failed the precision limits, of which14 were deemed to be of poor quality (clotted, brin clots, limited volume, etc.) on re-examination (see Table 2). One sample gave a result below the functional sensitivity of the assay. There remained four (0 08%) samples for which the cause of the CV% ctnl (mg/l) n CV% ctnl (mg/l) n CV% ctnl (mg/l) n CV% ctnl (mg/l) Figure 3. Precision of the Bayer ACS:180 assay for cardiac troponin I (ctni). Duplicate results were sorted into bins and the percentage coef cient of variation (%CV) calculated. Table 1. Impact of selection of the lower cut-off point on the consistency of risk classi cation for the assay of cardiac troponin I (ctni) on the Bayer ASC:180 Lower decision limit (mg/l) Number of samples for which either replicate is less than the decision limit Number of samples for which one replicate is on the opposite side of the decision limit than the other replicate Percentage misclassi cation* *Percentage of patients assigned to a different risk classi cation group (higher or lower) on repeat analysis of a given sample.

4 234 Sheehan et al. Table 2. The impact of sample quality on cardiac troponin I results on the ACS:180 Reagent Replicate Absolute lot no. First Second Mean %CV delta Comment after further investigation Obvious brin clots or particulate matter Very small sample with visible haemolysis Clotted sample Normal sample appearance Result below assay sensitivity Data are shown for those samples failing the duplicate precision limits (%CV420 and absolute delta40 16 mg/l). %CV ˆ percentage coef cient of variation. analytical discrepancy could not be explained. These results prompted us to reassess our samplehandling procedures to include a careful inspection of the sample both pre- and post-analysis. The Figure 4. The effect of small brin particles on the results of cardiac troponin I (ctni) on the Bayer ACS:180. A ne suspension of brin particles was added to a patient s sample; the sample was then divided into two aliquots, one of which was centrifuged prior to analysis (clear sample) and the other was mixed ( brin sample). impact of small brin particles on the troponin assay is illustrated in Fig. 4. The re-suspension of brin particles and cell debris in plasma samples prior to analysis is often an unintended consequence of attempting to visualiz e the sample prior to analysis and thus re ects day-to-day practice. Discussion The 4-month period during which all ctni measurements were performed in duplicate provided us with a unique opportunity to assess the impact of pre-analytical variables on the assay, as well as to con rm the assay s functional sensitivity, examine the precision of the assay and thereby the consistency of risk assessment at di erent decision limits under routine conditions. The recommendation from the ESC/ACC committee that a single cut-o set at the 99th centile of a healthy control population be used for troponin assays and that the required precision at this concentration should be 10% has obvious implications for the determination of the functional sensitivity of the assay. Our data indicate that the ctni concentration at which the inter-assay %CV is 10 for the Bayer ACS:180 assay is 0 27 mg/l; this does not support the use of the Bayer assay at a cut-o ctni concentration of 0 1 mg/l.

5 Decision limit for troponin I and assay performance 235 Although the manufacturer, as well as some publications in the literature, 3,8 have focused on the assay s MDC, we do not believe this is a useful or appropriate indicator of assay performance. The MDC is dependent on the composition of the zero calibrator in addition to assay performance and it may not re ect the precision achievable with material from patients. 9 We believe it is important to focus on the imprecision of the assay at the stated decision limit and the impact of this imprecision on the risk categorization of patients. If a ctni decision limit is used for assigning a patient to a particular risk group, it is important that clinicians have con dence in the risk assessment i.e. that repeat measurement on the same sample would not alter this assignment. Our results suggest that, at the decision limit of 0 1 mg/l, 7 7% of patients would be assigned to a di erent risk group on repeat analysis of the same sample due to analytical imprecision (Table 1). This includes both type I errors (i.e. reporting that an analyte is present when it is not) and type II errors (i.e. reporting that an analyte is not present when it is). 9 This is probably a conservative estimate given that the assays on each sample were run in duplicate in a single assay run. This intra-assay imprecision is likely to be lower than the inter-assay variation experienced in routine clinical practice since reagent, calibration and instrument variables are minimized. Therefore, day-to-day variability in risk assignment is likely to be even higher in practice and we doubt that clinicians would be satis ed with this performance. Reports in the literature 10^12 showing false-positive ctni results due to incompletely clotted samples and the requirement for rapid turnaround of results have led many laboratories to use heparinized plasma for this assay. Although only heparinized plasma samples were used for troponin measurement because of the known di erence between serum and plasma in ctni results on the ASC:180, 13 brin particles and cell debris in haemolysed samples were still a source of spurious results in our laboratory (n ˆ 14 samples). These represent routine samples that were deemed acceptable by the instrument operators on initial inspection and were detected because they exceeded duplicate precision limits. These results and the e ect of adding small brin particles to samples highlight the fundamental importance of presenting a sample of acceptable quality for analysis and led us to re-de ne our sample-processing procedures. All samples are now carefully inspected both before and after analysis. Although this study focused on the Bayer ACS:180 ctni assay, we believe that our results are applicable to other troponin assays and indeed other assays in general. In summary, we would suggest that, for the Bayer assay, and possibly for some other current-generation ctniassays, the use of a lower decision limit of 0 1 mg/l is inappropriate and can lead to misclassi cation in a small but signi cant number of cases. Although studies have shown that the assay can stratify groups of patients into low and higher risk at a troponin cuto of 0 1 mg/l, the impact of assay imprecision on the use of this cut-o in an individual patient cannot be ignored. We believe that there is a requirement for more-sensitive ctni assays to allow consistent individual risk assessment with troponins in the low abnormal range. This requirement is made more acute by the recent re-de nition of MI by the ESC/AACC. Newer assays, such as the Access 1 AccuTnI 2 and the Stratus CS, 14 demonstrate improved precision. We believe the stringent de nition of 510% imprecision at the 99th centile cut-o for MI is preferable to the 20% limit commonly applied to determine the lower immunoassay reporting range, given the consequences of incorrect diagnosis. This study is consistent with another recent study which showed that for a number of other commercial assays the functional sensitivity (at both10 and 20 inter-assay %CV) is higher than the cut-o determined by the upper reference limit of a control population. 15 Finally, we believe that laboratories need to work with clinicians to de ne assayspeci c decision limits that will support the clinical decision-making process. We recognize the merit in applying an MI cut-o at a predetermined concentration that meets the goal of 10% imprecision as suggested recently by others. 4 For the Bayer ACS:180 assay, our studies indicate that this cut-o should be 0 27 mg/l. References 1 Apple FS, Falahati A, Paulsen PR, Miller EA, Sharkey SW. Improved detection of minor ischaemic myocardial injury with measurement of serum cardiac troponin I. Clin Chem 1997; 43: Galvani M, Ottani F, Ferrini D, Ladenson JH, Destro A, Baccos D, et al. Prognostic in uence of elevated values of cardiac troponin I in patients with unstable angina. Circulation 1997; 95: Morrow DA, Antman EM, Tanasijevic M, Rifai N, de Lemos JA, McCabe CH, et al. Cardiac troponin I for strati cation of early outcomes and the ef cacy of enoxaparin in unstable angina: a TIMI- 11B substudy. J Am Coll Cardiol 2000; 36: Apple FS, Wu AH. Myocardial infarction rede ned: role of cardiac troponin testing [editorial]. Clin Chem 2001; 47: Wu AH, Apple FS, Gibler WB, Jesse RL, Warshaw MM, Valdes R Jr. National Academy of Clinical Biochemistry standards of laboratory practice: recommendations for the use of cardiac markers in coronary diseases. Clin Chem 1999; 45: Braunwald E, Antman EM, Beasley JW, Califf RM, Cheitlin MD, Hochman JS, et al. ACC/AHA guidelines of the management of patients with unstable angina and non-st segment elevation myocardial infarction: executive summary and recommendations. A report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines (Committee for the Management of Patients with Unstable Angina). Circulation 2000; 102:

6 236 Sheehan et al. 7 Alpert JS, Thygesen K, Antman E, Bassand JP, Apple FS, Armstrong PW, et al. Myocardial infarction rede ned: a consensus document of the Joint European Society of Cardiology/American College of Cardiology Committee for the Rede nition of Myocardial Infarction. J Am Coll Cardiol 2000; 36: Morrow DA, Rifai N, Tanasijevic MJ, Wybenga DR, de Lemos JA, Antman EM. Clinical ef cacy of three assays for cardiac troponin I for risk strati cation in acute coronary syndromes: a Thrombolysis in Myocardial Infarction (TIMI) 11B Substudy. Clin Chem 2000; 46: Davenport JM, Schlain B. Testing claimed minimal detectable concentrations of in vitro medical diagnostic devices. Clin Chem 2000; 46: Roberts WL, Calcote CB, De BK, Holmstrom V, Narlock C, Apple FS. Prevention of analytical false-positive increases of cardiac troponin I on the Stratus II analyser. Clin Chem 1997; 43: Nosanchuk JS. False increases of troponin I attributable to incomplete separation of serum. Clin Chem 1999; 45: Beyne P, Vigier JP, Bourgoin P, Vidaud M. Comparison of single and repeat centrifugation of blood specimens collected in BD evacuated blood collection tubes containing a clot activator for cardiac troponin I assay on the Access analyser. Clin Chem 2000; 46: Stiegler H, Fischer Y, Vazquez-Jimenez JF, Graf J, Filzmaier K, Fausten B, et al. Lower cardiac troponin T and I results in heparin plasma than in serum. Clin Chem 2000; 46: Altinier S, Mion M, Cappelletti A, Zaninotto M, Plebani M. Rapid measurement of cardiac markers on Stratus CS. Clin Chem 2000; 46: Yeo J, Kelly S, Quinn-Hall MT, Bateman SW, Fischer GA, Wieczorek S, et al. Functional sensitivity of cardiac troponin assays and its implications for risk strati cation for patients with acute coronary syndromes. In: Adams JE, Apple FS, Jaffe AS, Wu AH, eds. Markers in Cardiology: Current and Future Clinical Applications. American Heart Association monograph series. New York: Futura Publishing Co, 2001: 23-9 Accepted for publication 27 December 2001

Analytical performance of the i-stat cardiac troponin I assay

Clinica Chimica Acta 345 (2004) 123 127 www.elsevier.com/locate/clinchim Analytical performance of the i-stat cardiac troponin I assay Fred S. Apple a, *, MaryAnn M. Murakami a, Robert H. Christenson b,