One-Hour Troponin Using a High-Sensitivity Point-of-Care Assay in Emergency Primary Care: The OUT-POC Pilot Study

Acute chest pain is a common and usually benign symptom in emergency primary care [1]. In Norway, access to hospital emergency departments (EDs) generally requires being referred by a primary care physician or ambulance personnel [1]. At the same time, it is challenging to exclude an acute myocardial infarction (MI) in emergency primary care due to limited diagnostic decision tools [2‒4]. This often leads to defensive hospital referral practices and accumulation of low-risk patients in the EDs [3, 5].

In hospitals, clinical examination, ECG, and high-sensitivity cardiac troponins (hs-cTns) constitute the three diagnostic pillars for assessing patients with acute chest pain [6]. According to the European Society of Cardiology, the 0/1-h algorithm, comprising two hs-cTn measurements with a one-hour interval, is one of the preferred biomarker strategies [6]. However, hs-cTn testing faces logistical challenges in primary care due to its reliance on central laboratories, which in Norway are predominantly located in hospitals [7, 8].

In the observational One-hoUr Troponin in a low-prevalence population of Acute Coronary Syndrome (OUT-ACS) study, we demonstrated that the Roche Elecsys hs-cTnT 0/1-h algorithm was safe and applicable also among low-risk patients with chest pain in emergency primary care [9, 10]. With 77% of patients safely ruled out by the algorithm outside of hospital, the OUT-ACS strategy was also considered highly cost-effective [9‒11]. However, the generalisability of this approach has been limited to emergency primary care clinics close to a hospital with a 24/7 central laboratory access.

In the last decade, we have seen substantial improvements in point-of-care (POC) troponin assays, currently with four assays (LSI Medience PATHFAST; Siemens Atellica VTLi; QuidelOrtho TriageTrue; and SpinChip [SpinChip currently awaiting regulatory approval]) fulfilling the high-sensitivity criteria [12‒16]. In 2020, an assay-specific 0/1-h algorithm for the TriageTrue assay was derived and validated in a hospital cohort using frozen biobank samples with comparable safety when compared with the 0/1-h algorithms for the well-validated central laboratory assays hs-cTnI (ARCHITECT, Abbott Laboratories, IL, USA) and hs-cTnT (Elecsys, Roche Diagnostics, Rotkreuz, Switzerland) [13]. With these promising results, expanding the OUT-ACS strategy using a POC hs-cTn algorithm could now be possible. The TriageTrue hs-cTnI test may be a valuable decision tool in emergency primary care, including rural clinics. A nested sub-analysis from the Mersey Acute Coronary Syndrome Rule Out Study (MACROS)-2 randomised controlled trial confirmed the high analytical and clinical performance of the TriageTrue assay when using whole blood samples [17]. Still, no clinical studies have been published exploring the TriageTrue 0/1-h algorithm’s diagnostic performance when implemented in clinical care. Before considering initiating a larger multicentre study in Norwegian emergency primary care, we decided on a small-scale pilot study. The primary objective of this pilot study was to compare the TriageTrue 0/1-algorithm to standard care at our clinic (i.e., the 0/1-h algorithm for Roche Elecsys hs-cTnT) with respect to efficacy and to assess the feasibility and analytical precision of the POC hs-cTnI test when handled by nurses without laboratory expertise in emergency primary care.

The One-hoUr Troponin using a high-sensitivity Point-Of-Care assay in emergency primary care (OUT-POC) pilot study was a pragmatic, small-scale, prospective observational study. It was conducted between April 10 and June 13, 2024, at the Oslo Accident and Emergency Outpatient Clinic (OAEOC) in Oslo, Norway.

The OAEOC is the main emergency primary care clinic in Oslo. It is run by the City of Oslo municipality, serves the entire city 24/7 and has about 120,000 consultations annually. The clinic is staffed by general practitioners (GPs), nurses, and support personnel.

All patients presenting to the clinic as self-presenters or by ambulance are initially triaged by a nurse using the Manchester Triage System [18]. Patients with potential cardiac-suspected symptoms (i.e., chest pain/complaints, dyspnoea, fatigue, syncope, or palpitations) will have a 12-lead ECG recorded upon arrival and interpreted by the senior physician. If an ST-segment elevation MI (STEMI) is detected or the patient is considered critically ill, the patient will be directly hospitalised to prevent additional prehospital delay. A GP will then examine the patients according to their given triage. After clinical examination, the GP decides if supplementary troponin testing is needed for a conclusive decision. Patients with an apparent noncardiac condition (i.e., myalgia, anxiety, dyspepsia, pneumonia) are managed accordingly without prehospital troponin assessment. If a conclusive decision cannot be made, and the patient is considered not to be in urgent need of hospitalisation, serial troponin measurements are done at the clinic to rule-out acute myocardial injury/infarction (approximately 33% of the chest pain presenters) [19]. At the OAEOC, most of these patients stay in an observation unit while awaiting the results.

Patients ≥18 years of age presenting to the clinic with acute nontraumatic chest pain/complaints and a clinical indication for prehospital troponin measurements were consecutively approached for inclusion. Patients with a STEMI or considered to need direct hospitalisation were not available for study enrolment. Patients unable to provide written informed consent (i.e., language barriers or impaired cognitive function) were excluded. A sample size of 200 participants was preselected by the project group and considered sufficient for the purpose of this pilot study.

The troponin assessment routine comprises the collection of two venous blood samples, sampled at a one-hour interval. Blood sampling and study inclusion were performed by regular nurses at the clinic. During the pilot study, the samples were collected in three tubes (Greiner Bio-One, Kremsmünster, Austria): one 4 mL lithium heparin plasma tube (for the Roche Elecsys hs-cTnT analysis) and two 2 mL EDTA tubes (one sent to the central laboratory for other relevant analyses [e.g., haemoglobin], and one used for the hs-cTnI POC assessment if consent to study participation). The samples were continuously sent in pneumatic tubes for central laboratory centrifugation and analysis at the Department of Medical Biochemistry at Oslo University Hospital, Aker, located in the adjacent building.

Hs-cTnT was analysed on the Cobas 6000 e601 module analyser using the Elecsys Troponin T hs STAT assay (Roche Diagnostics, Switzerland). For hs-cTnT, the 99th percentile upper reference limit of a healthy reference population is 14 ng/L (with a coefficient of variation [CV] of <10%). The limit of detection is 5 ng/L, and the limit of blank is 3 ng/L [20, 21]. During the study period, the laboratory routinely analysed external quality assessment (EQA) material from Noklus (Bergen, Norway) with good performance, as well as internal quality assessment with a CV ≤8% at concentrations <15 ng/L and ≤6% at concentrations >15 ng/L.

The hs-cTnT results were interpreted by the treating GP, using the ESC 0/1-h algorithm [6] as decision support before discharge. By using assay-dependent thresholds, the algorithm triaged the patients to the rule-out group (i.e., low risk of acute MI), rule-in (high risk of MI), or further observation (Fig. 1a). The algorithm also includes specific single-sample rule-out criteria (i.e., hs-cTnT <5 ng/L and symptom onset ≥3 h) [6] and has been standard care at the clinic since April 2023 [19].

The regular nurses handled POC measurements and quality assessment of the POC MeterPro device (QuidelOrtho, San Diego, CA, USA) after receiving training and repeated information 2 weeks before study initiation. After each blood sampling, a standardised pipette (content of a TriageTrue High Sensitivity Troponin I Test kit) was used to extract 0.175 mL of whole blood from a 2 mL thoroughly mixed EDTA tube without further preparations, which was then transferred to a TriageTrue test cartridge before being inserted into the TriageTrue MeterPro system [22]. The hs-cTnI result was expected to be available in less than 20 min, and the automatically printed paper strips were stored for data analyses. During the pilot study, the treating GPs were blinded to the TriageTrue results, and the results were not made available in the patient records to avoid being mixed with the TnT results.

According to the manufacturer, TriageTrue has a 99th percentile upper reference limit of 20.5 ng/L with high analytical precision (CV 5.9–6.5% for whole blood). The limit of detection and limit of blank values are reported as 1.5–1.9 ng/L and 0.5–0.8 ng/L, respectively [16].

During the pilot study, internal quality assessment was performed once a week by the nurses by thawing and analysing frozen liquid controls delivered by QuidelOrtho (i.e., liquid controls based on EDTA human plasma containing preservatives and troponin I). A low or a high control was analysed every other week to (1) assess analytical reproducibility, and (2) confirm proper functioning of the cartridges stored at room temperature as determined by results being within the acceptance limits for the specific liquid control lot. This was a requirement to continue analysing patient samples. Serial analyses of the liquid controls by the same nurse on the same day were also performed to determine analytical repeatability, and all imprecision results were then reviewed at the Department of Medical Biochemistry at Oslo University Hospital.

After completed enrolment, the TriageTrue results were interpreted using the assay-dependent 0/1-h criteria, as published by Boeddinghaus et al. [13] Like the Elecsys hs-cTnT algorithm, the TriageTrue 0/1-h algorithm also includes single-sample rule-out criteria, further described in Figure 1b.

Clinical variables were collected from the electronic patient records. The collected variables were age, sex, diagnosis codes at discharge (from the International Classification of Diseases 10th Revision [ICD-10] [23]), further disposition (i.e., discharge home, follow-up by regular GP, referral to cardiac outpatient clinic, or acute hospitalisation), the initial ECG and a few clinical variables. Time variables were collected to assess the workflow, length of stay, and potential logistic challenges (including the onset of symptoms, arrival at the clinic, time to consultation and first blood draw, timing of samples, and final discharge). Hospital discharge diagnoses (ICD-10 codes) were registered from all patients sent to the hospital after the troponin assessment. All patient-sensitive data were electronically stored using a secure online server at the Norwegian Services for Sensitive Data platform by the University of Oslo [24].

Study participation also included registration of all ischaemic cardiac events (i.e., ICD-10 codes I20-I25) and cardiac-related deaths the following year through linkage with the Norwegian Cardiovascular Disease (CVD) Registry [25]. However, as these data will not be made available by the CVD Registry until Autumn 2025, they will be presented in a later publication.

The outcome measure for efficacy was triage discrepancies between the TriageTrue hs-cTnI POC 0/1-h algorithm and standard care (i.e., Elecsys 0/1-h algorithm for hs-cTnT). A discrepancy was considered major if a patient with acute MI had been triaged as rule-out by one of the algorithms and as rule-in by the other, while considered minor if the algorithms differed by one triage category (i.e., rule-out or rule-in to the observation group). The outcome measures for feasibility were assay- or turnaround times (TATs), and user- and device-related errors. The analytical precision outcomes reproducibility and repeatability were reported as CV.

All categorical variables are reported as numbers and percentages, and the continuous variables are reported as medians and interquartile ranges (IQRs). The pilot study was not powered to assess the performance (i.e., the sensitivity, specificity, and predictive values) of the two algorithms. Hence, these are not included in the analyses. The true performance will be evaluated during the next phase of the project. IBM SPSS version 29.0 (Armonk, NY, USA) was used for the data analyses. Microsoft Excel 2016 (Redmond, WA, USA) was used to calculate the standard deviations and CVs and to create the Levey-Jenning charts. The Sankey charts were generated through Sankey-MATIC.com.

Clinical personnel at the OAOEC (i.e., nurses and clinical personnel without laboratory technical expertise) were involved in the planning and preparing the pilot study, providing valuable feedback on the study material, data collection, and training before initiation. The personnel were also consulted to detect any challenges with the TriageTrue MeterPro system, test kits, and quality control routine and to evaluate workflow, pre- and analytical errors, potential process improvements, and timing and simultaneity conflicts.

A total of 199 patients were included with a median age of 54 years (IQR 45–70), and 105 (52.8%) were female (Table 1). The patients presented to the clinic 4.6 h (IQR 1.9–15.2) after the onset of symptoms and were assessed by a GP 56 min (IQR 32–96) after arrival. The initial ECG was interpreted as potentially ischaemic in 11 (5.5%) patients. Additional time variables and disposition after the final assessment are presented in Table 1. Among those hospitalised (n = 30, 15.1%), there were 5 patients with acute MI.

A total of 233 patients were approached for enrolment, and 34 patients (14.6%) were not included. Among these, there were 13 patients with language barriers, 6 patients declined, 6 patients were unable to provide informed consent, 8 patients were due to logistical challenges, and 1 patient had traumatic chest pain.

The first troponin sample was collected 8.0 (IQR 4.9–18.3) hours after the onset of symptoms. The 0-h Elecsys hs-cTnT sample triaged 63 (31.7%) patients towards direct rule-out, six (3.0%) to rule-in, and 130 (65.3%) patients required a second troponin sample after 1 h. On the contrary, by using the TriageTrue hs-cTnI POC assay, 129 (64.8%) patients were triaged towards direct rule-out, 8 (4.0%) patients to rule-in, and 62 (31.2%) patients required a second blood draw (Fig. 2).

The triage differences between the 0-h algorithms revealed minor discrepancies, i.e., a difference of one triage category, in 76 (38.2%) patients. As presented in Table 2, none of the patients with minor discrepancies in the POC rule-out group needed hospital transfer. There were no major discrepancies between the 0-h algorithms.

The second troponin sample was collected after 63 (IQR 61–67) minutes. To avoid missing the 1-h window, the nurses at the clinic always collect the 1-h sample from patients where troponin measurements are requested [19]. By adding the 1-h results, the hs-cTnT 0/1-h algorithm triaged 148 (74.4%) patients as rule-out, 13 (6.5%) as rule-in, and 38 (19.1%) remained in the observation group. Similar results were found for the hs-cTnI POC 0/1-h algorithm, triaging 146 (73.4%) as rule-out, nine (4.5%) as rule-in, and 44 (22.1%) to the observation group (Fig. 2).

After adding the 1-h measurement, 158 (79.4%) patients ended up in the same triage category by the two algorithms. Minor discrepancies were detected for 36 (18.1%) patients, while five (2.5%) had major discrepancies (i.e., triaged towards rule-out by one of the algorithms and rule-in by the other). None of the four POC rule-out cases, triaged as rule-in by the hs-cTnT algorithm, presented any apparent safety issue for the POC assay. Additional clinical features in the patients with major discrepancies are presented in Table 3.

The median analysis time for the TriageTrue assay, from cartridge insertion to the printed result, was 13 min (IQR 12–15). For the Elecsys hs-cTnT assay, analysed at the central laboratory, the median TAT (i.e., from blood sampling to reported result) was 48 min (IQR 38–67; with reported 90 percentile of 92 min as a laboratory quality indicator).

Among 398 potential POC measurements (199 patients 2x), errors occurred only 8 times (2.0%). Six of these were directly related to staff errors (i.e., 2x EDTA tubes not being used, 2x cartridges inserted too late into the device due to time conflicts, 1x sample forgotten, and 1x too small volume transferred to the cartridge). Hence, device-related error codes were as low as 0.5% (n = 2).

During the pilot study, 24 liquid TriageTrue hs-cTnI quality controls originating from two different lots were analysed by the nurses at the clinic. All were well within the acceptance limits and thus approved. For the analytical reproducibility (controls analysed every other week by several nurses), we found a CV of 8.9% for target 27.0 ng/L (acceptance limits 15.0–39.0 ng/L) and a CV of 6.1% for target 557 ng/L (acceptance limits 306–808 ng/L) (Fig. 3a, b). For the analytical repeatability (controls analysed on the same day by the same nurse), we found a CV of 6.6% for target 29.0 ng/L (acceptance limits 16.0–42.0 ng/L) and a CV of 8.6% for target 513 ng/L (acceptance limits 282–744 ng/L) (Fig. 3c, d).

In this pragmatic, small-scaled pilot study, the efficacy, feasibility, and robustness of the TriageTrue hs-cTnI POC assay were assessed when used by non-laboratory personnel in emergency primary care. There were four main findings: firstly, after a single blood draw, the TriageTrue POC algorithm was demonstrated to be far more effective than standard care (i.e., the Elecsys hs-cTnT assay), with a rule-out proportion of 64.8% versus 31.7%, respectively. The rule-out efficacy increased to 73.4% when adding a 1-h measurement, comparable to the hs-cTnT assay (74.4%). Secondly, no critical discrepancies were found when comparing the two algorithms, nor were there any apparent safety issues for patients triaged to the POC rule-out group.

Thirdly, the TriageTrue POC hs-cTnI median analysis time was 13 min. Although the median TAT (i.e., TAT, from sampling to reported result) at the central laboratory was 48 min, this is often followed by delayed evaluation by the treating GP. Therefore, most patients at the clinic will routinely have a second hs-cTnT measurement collected to avoid missing the 1-h window [19]. Implementing the TriageTrue POC assay could prevent a second blood draw in 7 of 10 cases, reducing both unnecessary testing and delayed clinical decisions.

Finally, the POC assay demonstrated high robustness and user-friendliness, as the analytical quality assessment, performed independently by non-laboratory personnel, achieved satisfactory results. Both reproducibility and repeatability results were adequate, and device-related error codes occurred in only 0.5% of the samples.

Implementing a rapid POC 0/1-h algorithm in emergency primary care is expected to aid clinical decisions, resulting in more patients being discharged home. This pilot study was launched to address some uncertainties before proceeding with a more extensive study, as the scientific documentation of the TriageTrue 0/1-h algorithm at the time was limited to the publication by Boeddinghaus et al. [13] Although excellent safety was presented, their validation study was based on frozen plasma samples, known to have better stability than whole blood samples [26]. Since then, the MACROS-2 trial has been published, evaluating the TriageTrue assay using whole blood samples from an ED setting [17]. Similar to our findings, they report a high proportion of single-sample rule-out cases (61.6%), high analytical stability for whole blood samples, and few device-related errors when handled by non-laboratory staff [17].

Single-sample rule-out thresholds have also been derived and validated for the Siemens Atellica VTLi hs-cTnI POC assay for ED settings using archived samples. In the US SEIGE (Safe Emergency Department Discharge Rate) derivation cohort (MI prevalence 8.1%), the rule-out proportion was as low as 17.8% [27], increasing to 41.8% when applied in the Australian Suspected Acute Myocardial Infarction in Emergency (SAMIE) validation cohort with lower MI prevalence (5.5%) [27]. The VTLi rule-out efficacy was later increased to 53.3% and 66.1% in the two cohorts when deriving and validating a novel 0/2-h algorithm [28]. One of the main purposes of introducing hs-cTn POC testing in emergency primary care is to prevent unnecessary hospital referrals, so a safe and reliable POC assay identifying a high proportion of low-risk cases is highly needed. Subsequently, the promising feasibility and its high proportion of single-sample rule-out cases make the TriageTrue algorithm especially attractive for use in a low-risk primary care setting.

This is the first study to present clinical prospective data on the TriageTrue 0/1-h algorithm conducted by end-users in a non-laboratory, low-prevalence setting. With the robust findings from the MACROS-2-trial [17] and the knowledge learned through our pilot study, proceeding with a more extensive implementation study using the TriageTrue POC assay in emergency primary care now seems appropriate.

Some limitations need to be addressed: firstly, the pilot study was not powered to assess the predictive performance of the algorithms, as the number of patients is low, and the few MI cases registered from hospital discharge documents have not been adjudicated. Supportive registry data will not be available until Autumn 2025. Hence, without information on any MIs or deaths among those sent home, we cannot comment on the rule-out safety (i.e., the sensitivity or negative predictive values), nor how the TriageTrue algorithm will perform in subgroups (i.e., early presenters). These essential questions will be addressed in the following phase of the project.

Secondly, in this pilot study, we aimed to compare the efficacies of the two 0/1-h algorithms. However, a direct comparison of the TnI and TnT values was not possible, as each hs-cTn assay will have its analytical sensitivity and assay-dependent thresholds [26, 29]. However, comparing the two algorithms’ ability to categorise patients into rule-out, observation, and rule-in, which indicate a low, moderate or high probability of acute MI, are clinically more relevant. Hence, we settled for a pragmatic comparison of the algorithms and their clinical triage categories for this study without detecting any apparent safety issues for the whole blood POC assay.

Thirdly, the decision to perform troponin testing is made by the treating GP, which introduces a selection bias. Although our participants do not represent the whole chest pain cohort at the clinic, we argue that this pre-selection is more of a strength than a limitation. Screening all chest pain patients for troponin levels at presentation in this low-prevalence setting would cause an extensive overuse of healthcare resources. Hence, our study focuses on a selected population that we believe would benefit from additional testing in primary care.

Fourthly, a direct comparison of the TATs of the two methods was not feasible. The central laboratory samples were prioritised as the POC results were blinded to the GP. Consequently, the timeline includes blood sampling, preparation, and transport of samples to the hospital laboratory for hs-cTnT analysis before running the whole blood test by POC. This would not reflect an accurate TAT for the POC test. Therefore, we report the timeframes we could measure: the TAT for Elecsys hs-cTnT assay and the analysis time (from cartridge insertion to reported results) for the TriageTrue POC hs-cTnI assay.

Finally, an internal quality assessment was performed weekly with good imprecision results using material provided by the manufacturer. Unfortunately, we could not conduct a sponsor-independent evaluation of trueness as no TnI external quality material existed for the EDTA-based MeterPro system during the pilot study. However, we expect this to be available soon and hopefully before initiating the next phase of this project.

In this small-scaled, pragmatic pilot study, the novel 0/1-h algorithm for the TriageTrue hs-cTnI POC assay appears efficient, feasible, and robust when applied in a non-laboratory-based low-risk setting outside the hospital. Also, with the high proportion of single-sample rule-out cases, repeated 1-h measurements could likely be avoided in the large majority of cases, enhancing patient management with decreased length of stay and healthcare expenditure. Further studies are needed to investigate how this POC algorithm may enhance clinical decisions in emergency primary care.

The authors would like to thank the patients participating in this study, Dr. Sven Eirik Ruud, Hege Karlsen, Linda Teigen, Gjermund Nilsen, and the dedicated personnel at the OAEOC Observation Unit. A special appreciation goes to Dr. Beate Nilsen, Head of the Section of Emergency Primary Care at the OAEOC, for always encouraging and facilitating clinical research at the clinic. The authors would also like to express their gratitude to QuidelOrtho’s Anne Noer, Ellen-Margrete Wahl, and especially Dr. Annemieke Döhmann for their continuous and thorough support throughout the pilot study.

Written informed consent was obtained from all participants. In accordance with the Helsinki Declaration, the pilot study was reviewed and approved by the Regional Committee for Medical and Health Research Ethics in Norway (ref. 629727), the Information Security and Privacy Office at the University of Oslo (ref. 828485), and the Information Security and Privacy Office at the City of Oslo Health Agency.

T.R.J. and D.A. have participated in webinars hosted by QuidelOrtho during 2023-24 without receiving any financial compensation for their participation. D.A. has received speaker fees from Amgen, Amarin, AstraZeneca, Bayer, Boehringer Ingelheim, BMS, GSK, MSD, Novartis, Novo Nordisk, Pfizer, Pharmacosmos, Philips, Roche Diagnostics, Sanofi, Takeda, and Vifor, unrelated to the present work, and grant support (to the Institution) from BMS/Pfizer, Medtronic, Bayer, and Roche Diagnostics. SH has received speakers’ honoraria from AstraZeneca, Bayer, Boehringer Ingelheim, Bristol-Myers Squibb, Sanofi, Novartis, and Novo Nordisk, unrelated to the present work. DA was a member of the Journal’s Editorial Board at the time of submission. The remaining authors have no conflict of interest to declare.

T.R.J. received a research grant from the Norwegian Committee on Research in General Practice. The remaining authors received no funding. QuidelOrtho Corporation, California, USA, supported the study with POC devices, testing material, and personnel training before study initiation. The pilot study is part of an investigator-initiated project. Hence, the funders had no role in data acquisition, interpretation, or presentation of the results.

The study was conceived and designed by T.R.J. in close collaboration with O.M.V., A.C.K.L., S.H. and D.A., T.R.J. and A.C.K.L., with support from QuidelOrtho, trained the personnel before study initiation. T.R.J. oversaw data collection, compiled the data, performed statistical analyses, and drafted the manuscript, with contributions from the co-authors. A.C.K.L. created the Levey-Jennings control charts. The project team has reviewed, revised, and approved the final version for submission.

The data supporting this study are not publicly available due to a planned future publication. However, raw data can be requested from the corresponding author (T.R.J.) upon reasonable request.