The underestimated issue of non-reproducible cardiac troponin I and T results: case series and systematic review of the literature

Case report 1

On May 2019, an 80-year-old female went to her general practitioner (GP) to get a blood test for diabetes control. The patient is known with heart failure, hypertension, hypercholesterolemia, and hypertriglyceridemia and complained about chronical retrosternal pain since several months. The patient was sent to our emergency department (ED) based on an increased high-sensitivity cardiac troponin T (hs-cTnT) concentration (102.0 ng/L; 99th upper reference limit [RL] = 14.0 ng/L; Roche Diagnostics, Basel, Switzerland). Creatine kinase-MB isoenzyme (CK-MB, immunological; Roche Diagnostics) and N-terminal pro B-type natriuretic peptide (NT-proBNP; Roche Diagnostics) were 3.4 μg/L (RL <4.9 μg/L) and 557 pg/mL (RL <486 pg/mL), respectively. Fifteen years ago, the patient was diagnosed with transient ischemic attack. Her familial history revealed fatal MI by her father, five uncles and aunts. Her brother also died from stroke. A new blood sample was drawn in the ED 5 h after the first one, and a significant hs-cTnT drop was observed (15.1 ng/L). Based on this large difference, the first sample drawn by the GP was re-assayed in our laboratory and near normal hs-cTnT result was obtained (14.4 ng/L). The sample drawn in the ED was also re-assayed and gave somewhat similar results (14.0 ng/L). Electrocardiogram (ECG) was performed and was characterized with left bundle branch block. The patient was admitted to the cardiology service where a coronary angiography was performed and a pharmacological stent was placed. The first result (102.0 ng/L) was considered to be a false positive and non-reproducible result.

Case report 2

On December 2019, a 53-year-old man was seen by his GP for a follow-up visit, owing to slight chest discomfort. The patient was known with hypercholesterolemia and hypertension. Based on the clinical presentation, the GP ordered a cTn test. A few hours later, our laboratory called the GP because of the elevated hs-cTnT (84.0 ng/L, 99th percentile of the assay = 14.0 ng/L). Although creatine kinase (CK) and CK-MB were both elevated (301 U/L; RL <190 and 7.7 μg/L; RL <4.9 μg/L, respectively), the CK-MB index (CK-MB*100/CK) was negative (2.6; RL <3.5). In light of this information, the GP decided to refer the patient to the ED. There, a new sample was taken and a normal hs-cTnT result was obtained (6.8 ng/L). CK-MB was still increased (8.6 μg/L) and the CK-MB index was still negative (2.7). The first sample was subsequently measured again, which returned a normal hs-cTnT result (5.3 ng/L). The patient’s ECG, echocardiography, and stress test were all normal. For this patient, hs-cTnT was, thus, falsely elevated. Fortunately, the incorrect result did not lead to any unneeded invasive procedures [2].

Case report 3

On December 2019, a 65-year-old man with known chronic obstructive pulmonary disease was hospitalized for a respiratory tract infection. C-Reactive Protein (CRP) and neutrophils count were both elevated (224 mg/L; RL <5 mg/L and 11.0×10³/µL; reference interval 2.0–7.5×10³/µL, respectively). The patient’s hs-cTnT was also substantially elevated (393.0 ng/L). Three hours later, a new sample was taken, and normal results were obtained (10.3 ng/L). Both samples were measured again, which returned normal results (12.0 and 13.0 ng/L, respectively). CK-MB and NT-proBNP were both normal, and the ECG revealed no abnormalities. For this patient, hs-cTnT was thus again falsely elevated. Fortunately, the incorrect result did not lead to any unneeded invasive procedures [2].

Case report 4

On September 2020, a 43-year-old man was seen by his GP for intermittent chest pain. The patient is known with Behçet’s disease and takes ciclosporin. The patient was sent to our ED based on an increased hs-cTnT concentration (42.3 ng/L) performed at our laboratory. However, ECG and echocardiography investigations were both normal in this patient. A new blood sample was drawn in the ED 4 h after the first one, and a significant drop in the measured hs-cTnT level was observed (10.8 ng/L). Based on this large difference, the first sample drawn by the GP was re-assayed in our laboratory and a normal hs-cTnT result was obtained (11.8 ng/L). The sample drawn in ED was also re-assayed and gave similar results (12.0 ng/L).

Based on the observation of these discordant results, we checked the precision of our hs-cTnT assay at several occasions. Using patient pools, coefficient of variations near the decision level (i.e., 14 ng/L) ranged from 2.3 to 3.9%. Roche Diagnostics also checked our analytical process and did not identify any failure. Of note, the four cases abovementioned occurred on three different reagent lots.

Interferences in troponin assays

False positive cTn results have been related to the presence of anti-cTn antibodies or immunocomplexes [3], [4], heterophilic antibodies [5], elevated alkaline phosphatase level [6], contrast media [7], instrument malfunction [8], or carry-over [9]. False negative results with cTn assay may also be observed in the case of a large biotin intake [10], [11]. All these interferences are characterized with erroneous but reproducible cTn results: in other words, if the sample is reanalyzed a second time, the same erroneous result will occur [12], [13]. Non-reproducible (also named irregular) interferences have also been reported in the literature [14]: in this case, if the sample is reanalyzed a second time, a significant difference in cTn results will be observed. These interferences have been described as “fliers” or “outliers” [14]. In this review, we decided to use the term “outlier”. It is well known that the irregular analytical errors due to interferences in immunoassay systems are not only variable and sporadic but also clinically misleading because they can lead to both false-positive and false-negative results [14]. The aim of this review is to warn clinicians and medical laboratory scientists about these non-reproducible results that may alter the clinical judgment. Moreover, we aimed at identifying the nature of these non-reproducible cTn results, determining their occurrence, and describing the best way for their identification.

Identification through recentrifugation

Studies that used the recentrifugation method to identify outliers are present in Table 1. Grossly, the outlier rate varied from 0.17 to 28.57% in serum and from 0.091 to 4.0% in plasma. Outlier definition varied widely between studies. For example, Li et al. described that a difference of 30% was sufficient [31], while a difference of two SD was required for Fleming et al. [30]. Some studies did not report a clear outlier definition [3], [19], [20]. Hejl et al. defined that an original result above 100 ng/L and below 100 ng/L after recentrifugation was a false positive result [32]. However, this approach does not take into account assay imprecision, and so the number of false positive samples may be grossly overestimated [33]. The integration of assay imprecision in the outlier definition is therefore, in our opinion, critical. Inclusion criteria were also different across studies. Fleming et al. only included samples with cTnI >100 ng/L along with normal CK [30], while Beyne et al. included all range of cTnI concentrations [34] and Warner et al. [3] samples with a discordance with another platform (Ortho Vitros). Noteworthy, Kazmierczak et al. noticed that the occurrence of these non-reproducible results was not specific to a particular reagent lot [22]. Sheenhan et al. also found that two distinct reagent lots were both impacted by outliers [20].

Considering the large difference in analytical performances and measured results among hs-cTnI and hs-cTnT methods [35], [36], it is not surprising that a large dishomogeneity in outlier frequency have been reported in the literature. Of interest, Lee et al. identified a higher outlier rate with a cTnI assay (0.97%) compared to a hs-cTnI assay (0.09%) [27]. More recently, Karon et al. confirmed that a hs-cTnI assay had a lower outlier rate (0.47%) compared to cTnI assay (3.66%). However, the outlier rate of the hs-cTnI assay was not statistically different compared to a cTnT assay [26]. The use of fixed or sex-specific cut-offs modified the number of outliers [26]. If we focus our analysis on studies published since 2010 that used one of the three main outlier definitions, the outlier prevalence is overall lower in high-sensitive assays compared to previous generation assays (Table 3).

Identification through repeat testing

Studies that used the repeat testing method (i.e., without recentrifugation) to identify outliers are presented in Table 2. Grossly, the outlier rate varied from 0.0 to 34.48% in serum and from 0.10 to 2.0% in plasma. Included numbers of samples were far higher as compared to studies having used recentrifugation. Favresse et al. included up to 1,243 samples [2], Pretorius et al. 2,391 samples [23], Pretorius et al. 4,336 samples [29], Ryan et al. up to 4,009 samples [24], Sheehan et al. 4,850 samples [20], Ungerer et al. up to 5,265 samples [12], and Sawyer et al. up to 7,089 [13].

Inclusion criteria and outlier definition varied also greatly across studies. The z value definition was used in the study of Pretorius et al. [23], [29], while Ungerer et al. [12], Favresse et al. [2], Ryan et al. [24], and Sawyer et al. [13] used the CD definition. Saywer et al. found that the hs-cTnI assay was less affected than contemporary cTnI assays for critical outliers [13]. Using the hs-cTnT method, Pretorius et al. found two outliers above the 99th percentile cut-off out of 2,391 duplicate samples (outlier rate = 0.08%; z>3.48) [23]. This outlier rate was significantly lower compared to previous generation assays [23]. However, Ryan et al. found that the outlier rate was not different between the hs-cTnI assay and the contemporary cTnI assay they previously used [24]. In studies published since 2010 and using one of the three main outlier definitions, the outlier prevalence is overall lower in high-sensitive assays compared to previous generation assays (Table 3). Interestingly, Sawyer et al. showed that a lower outlier rate was observed following a full maintenance of their platform [13]. Additionally, Kavsak et al. suggested that other factors contributing to variability may include periods of inactivity of the analyzer as well as the reagent pack sizes [37]. No outlier was classified as a critical outlier in their study.

In our recent report, we showed that the outlier rate varied among two reagent lots used (hs-cTnT). An outlier rate of 3.22 and of 0% was observed on two different and consecutive reagent lots (429178 and 460113). The high outlier rate observed on the first reagent lot was due to contamination with magnetic/paramagnetic particles [2]. Further using the lot 460113 between March 28 and June 6, 2020 (1,998 samples requests for cTn), we encountered two outliers (19.6 to 48.5 and 67.7 to 4.4 ng/L) out of 999 samples analyzed in duplicate. Therefore, the updated outlier rate was 0.093% (4/2,159) and the release of the new reagent lot drastically decreases the occurrence of outliers (Figure 1). On June 7, we installed a new reagent lot (470034). Between June 7 and September 23, we analyzed 1,438 cTn samples in duplicate and four outliers were observed (17.0 to 4.0, 39.8 to 22.7, 42.3 to 10.8, and 28.0 to 3.7 ng/L), leading to an outlier rate of 0.28% (Figure 1). On October 13, we installed a new reagent lot (474941). Between October 13 and December 16, we analyzed 1,748 cTn samples in duplicate and two outliers were observed (25.6 to 6.4 and 15.8 to 6.5 ng/L), leading to an outlier rate of 0.11% (Figure 1). These latter outlier rates were close to the one published by Pretorius et al. using a hs-cTnT assay [23].

Figure 1:

Rates of outliers observed in Clinique Saint-Luc Bouge using four different hs-cTnT reagent lots.

Dark blue = rate of outliers calculated by the CD formula; purple = rate of critical outliers (i.e., first cTn result > cut-off and second cTn result < cut-off).

Discussion

Outliers have been described in all cTn methods and it is strictly related to assay performance. Outliers occur randomly and can produce both false positive and negative results (14). The outlier rate vary according to the outlier definition used (i.e., critical outlier, z value [z=3.29, 3.48, …]), to the samples included (i.e., random samples, positive cTn results, samples with negative CK-MB results, serum vs. plasma, storage conditions, delay before reanalysis), to the procedure used to identify outliers (i.e., repeat testing or recentrifugation) and to the platform considered (including reagent lots). The direct comparison of studies is therefore difficult. The use of a repeat testing strategy in clinical practice is more convenient than recentrifugation (extra manipulation and TAT). Furthermore, recentrifugation might have an impact on the measurand as warned by the Clinical and Laboratory Standard Institute. Indeed, Canovi et al. showed a median reduction of 10.0 ng/L following recentrifugation [38]. The stability of cTn during a second centrifugation process and a possible increased adherence to the container cannot be excluded [33]. Third, some studies show that recentrifugation had no added value compared to repeat testing to identify outliers [12], [23].

We also focused our analysis on recent studies (i.e., published since 2010) to assess the outlier prevalence of previous and new generation assays (i.e., high-sensitive assays). The outlier definitions included were the most used ones (exceeding a z score value, CD or critical outlier). The outlier prevalence is overall lower in high-sensitive assays (0.28–0.57%) compared to previous generation assays (0.33–1.65%) (Table 3). In contrast to outliers, the biotin interference in cTn assays has received much more attention in the literature since the U.S. Food and Drug Administration notified a death attributable to a false rule-out of MI based on a falsely decreased cTn result. However, the prevalence of biotin >20.0 ng/mL in real-life studies was found to be 0% (0/572) and 0.13% (1/797) [39], [40]. Furthermore, the misclassification risk due to biotin interference was even lower (0.025, 0.0064, 0.00048%, and <0.00001% at 0, 1, 3, and 6 h) [40]. The prevalence of cTn outliers is therefore higher (Table 3). To limit the impact of biotin in cTn assays, manufacturers have added blockers that successfully overcame the interference up to 1,200 ng/mL [11]. The prevalence of heterophilic antibodies in cTn assays has been found to range from 0.07 to 3.0% [21], [41] and manufacturers also add blockers to limit interference due to heterophilic antibodies [5]. The identification of outliers is however more problematic.

A pragmatic solution would be to test every elevated cTn in duplicate to identify outliers. However, such a solution would be costly and significantly increases the TAT. In our laboratory (Saint-Luc Bouge) the prevalence of positive cTn results is about 50%. Testing in duplicate would hence double the reagent cost. In the same way, we suggest that a 1-h (or 3-h) cTn result significantly lower (>30% difference [42]) than the 0-h result (0-h/1-h [or 3-h] algorithm) could be investigated for a potential outlier in patients presenting with symptoms suggestive of acute MI (including acute chest pain and angina pectoris) [43], [44]. Automatic comments and repeat testing could therefore be generated in the laboratory information system in case of particular cTn kinetics. Continuous discussion between clinicians and the laboratory is fundamental to identify these erroneous and non-reproducible results. We further recommend that manufacturers continuously improve the robustness of their procedures to avoid erroneous cTn results that may potentially lead to harmful consequences for the patient. Given the growing data on the subject, in addition to precision and trueness, the frequency of analytical outliers should be systematically evaluated before the implementation of a new cTn assay. We therefore also encourage manufacturers, in collaboration with clinical laboratories and regulatory bodies, to conduct large studies to assess the occurrence of these outliers and to provide the probability of outliers using their assays. These evaluations could be performed for cTn but also for other measurands for which outliers have also been identified [45].