The effect of multiple analysers on the biochemical diagnosis of myocardial infarction using a contemporary troponin–I assay

Introduction

An accurate and precise measurement of troponin is crucial for the diagnosis of myocardial infarction (MI). The diagnosis requires serial measurements of troponin between 3 h and 12 h that show a significant rise and/or a fall in troponin values, with at least one measurement above the diagnostic threshold.^1,2 It is also recommended that laboratories define significant change between serial troponin measurements. ³

Troponin assays are expected to have an analytical coefficient of variation (CV_A) of ≤10% at the diagnostic threshold, although assays with a CV_A between 10% and 20% are clinically acceptable. ² The high-sensitivity troponin (hsTn) assays achieve coefficients of variation of <10% at troponin concentrations below the diagnostic threshold. At these concentrations, contemporary troponin assays have higher imprecision.^4,5 Although hsTn assays are available in mainland Europe and the United Kingdom, the use of contemporary assays remains widespread. Currently around 78% of users in the United Kingdom use contemporary troponin assays (based on recent United Kingdom National External Quality Assessment Service (UK NEQAS) data). In addition, the hsTn assays are not available in the USA and await Food and Drug Administration (FDA) approval.

Many laboratories utilize two or more analysers for the measurement of troponin, and samples are randomly distributed by automation to any analyser. It is therefore important to consider the impact of the combined analytical performance when multiple analysers are used upon serial troponin measurement, especially where contemporary troponin assays are utilized. In this study, we investigated the performance of a contemporary troponin-I (cTn-I) assay, operated across three analysers, connected to an automated track and sample manager, in order to established reference change values (RCVs).

Method

Following routine cTn-I analysis, serum samples that had been stored at 4–8℃ for <24 h were anonymously pooled to produce samples with target cTn-I concentrations between 20 ng/L and >500 ng/L. Aliquots of each pool were prepared and stored frozen at −20℃ until analysis. They were defrosted once and front loaded onto one of three ADVIA Centaur XP (XP1, XP2 and XP3) instruments (Siemens, Erlangen, Germany) for analysis with a contemporary^4,5 TnI-Ultra assay. The diagnostic threshold (where CV_A ≤ 10%) and the limit of detection are reported as 40 ng/L and 6 ng/L, respectively.^4,5 One of the above pools, therefore, had a target cTn-I concentration of 40 ng/L, the diagnostic threshold. The analysers were programmed to measure 200 μL aliquots, with different cTn-I concentrations, simultaneously at variable times on different days for 10 days. All runs were performed using the same reagent lot number. The results obtained on running internal quality control and external quality assurance samples, at the time of each sample measurement, were confirmed to be within acceptance. A total of 210 inter-batch data points were obtained. All statistical analyses were performed using R Project for Statistical Computing.

The statistical difference between the mean values at each concentration was computed by analysis of variance (ANOVA). In addition, performance in terms of accuracy of the three analysers was compared across the concentration range by analysis of covariance (ANCOVA). Overall performance of the three analysers was ascertained by considering them as a single unit and aggregating the data at each target troponin concentration. Precisions plots (reciprocal model) were generated for the aggregated CV_A values and used to predict inter-assay CV_As for the three analysers combined.

RCV was calculated using the formula RCV = 2.77√( CV I 2 +  CV A 2 ) (where CV_I = intra individual variation, taken as 10% ⁶ and CV_A = aggregate inter-batch analytical CV_A). ⁷

Results

CV_As improved with increasing cTn-I concentration and ranged from 22.1% to 5.8% at target cTn-I concentrations 20–500 ng/L. This gave RCVs of 67.2% (±13 ng/L) to 32% (±160 ng/L) across this concentration range. At the diagnostic threshold of 40 ng/L, the CV_A of the aggregated data for all three analysers was 13.6%. RCV was therefore estimated as 46.8% (±19 ng/L) at the diagnostic threshold (Figure 1). Figure 1 also shows that the CV_A and RCV are somewhat consistent at cTn-I concentrations over 100 ng/L). With aggregated mean value of 38 ng/L (target: cTn-I 40, range: 29.5–54.5 ng/L), the overall proportion of cTn-I measured at >40 ng/L was 0.33. Broken down by analyser (XP1, XP2, XP3), the proportions were 0.00, 0.50 and 0.50, respectively (H₀ proportions are equal, P < 0.05) showing significant disagreement between XP1 and the other two analysers (Figure 2). OPEN IN VIEWER

OPEN IN VIEWER

Figure 2.

Scatter plot of cTn-I measurement on each analyser at approximate target concentration 40 ng/L. cTn-I: contemporary troponin-I.

Discussion and conclusions

This study demonstrates that, in terms of bias, differences in the performance of different analysers has clinical implications at the diagnostic threshold. The differences in cTn-I measurement when multiple analysers are used can generate false positive or false negative biochemical criteria for the diagnosis of MI. Despite this, the combined CV_A across our three analysers at the diagnostic threshold is 13.6%, within the clinically acceptable range for cTn-I assays.

Laboratories are expected to report significant change values between serial troponin measurements, ³ but there is no clear consensus on best practice for this. There is debate on whether to report a single or multiple RCVs for troponin measurement and whether to report absolute or percentage change. ⁸

We determined a range of RCVs between 67.2% and 32% for the cTn-I concentrations we investigated for our three combined analysers. We therefore propose that reporting a single RCV as a measure of significant change for cTn-I may not be appropriate at cTn-I concentrations below 100 ng/L in routine practice, in those laboratories where multiple automated analysers are used.

We believe this to be the first report, of the impact, on analytical performance, of running a cTn-I assay simultaneously on multiple analyser platforms. Our findings should be of interest to the many routine service laboratories, which like our own, operate a cTn-I assay on multiple platforms.