Does high-sensitivity troponin measurement aid in the diagnosis of pulmonary embolism?


Kerstin Hogg, Clinical Lecturer in Emergency Medicine, 3rd Floor Clinical Sciences Building, Salford Royal Hospital, Stott Lane, Manchester M6 8HD, UK.
Tel.:+44 161 286 2334; fax: +44 161 206 4345.

Venous thromboembolism (VTE) affects around one in 1000 of the population every year [1]. Assessment for pulmonary embolism (PE) requires a sensitive and specific tool, to ensure timely treatment for those with PE, without false-positive diagnoses and unnecessary anticoagulation.

D-dimer testing in combination with pretest probability excludes PE in up to half of ambulant outpatients [2]. Blood D-dimer is sensitive for PE but cannot be used without an estimate of patient pretest probability. On occasion, physicians interpret D-dimer values in isolation or fail to perform a probability assessment [3,4]. Recent work has focused on cardiac troponin in PE prognosis. One meta-analysis [5] analyzing troponin levels in hemodynamically stable patients with PE reported that 26% of patients had abnormally high troponin T or I levels on diagnosis. The studies used troponin assays which measure as low as 0.01 ng mL−1, any detectable level being considered as abnormally elevated. Roche Diagnostics have a new fifth generation troponin T assay, which has a lower limit of detection of 0.003 ng mL−1. This assay will measure levels considered to be in the ‘normal’ range [6].

The Thromboembolism Assessment and Diagnosis (THREAD) study hypothesis was that the new generation troponin T assay might identify patients who have PE. Our aim was to assess the role of this assay in the diagnosis of PE.

The study was approved by the Local Ethics Committee and conducted in a 900-bed university hospital. Consecutive patients suspected of either deep vein thrombosis (DVT) or PE were recruited; however, this analysis is restricted to those investigated for PE. Any patient investigated as an inpatient or outpatient was eligible. Exclusions were inability or refusal to consent, and age under 16 years.

The first routine biochemistry blood sample taken at presentation of symptoms was stored in a secure −80° freezer, for further analysis. Samples were analyzed by a biochemist, blinded to patient diagnoses. Troponin T was measured by immunoassay using the fifth generation kit (Roche Diagnostics, Burgess Hill, UK) on the E170 Modular analyzer. The assay was calibrated and quality controlled according to the manufacturer’s guidelines. Precision was found to be acceptable with between batch percentage CVs being 2.4 and 2.3 at levels of 0.114 and 3.966 ng mL−1, respectively.

Each patient underwent standardized assessment for PE. Patients’ scoring low Wells’ probability (< 2.0 points) with a negative latex agglutination D-dimer test (IL Test D-dimer, cutoff < 230 ng mL−1 according to the manufacturer’s guidelines) had PE excluded. Pregnant patients and injecting drug users had diagnostic imaging, regardless of their Wells’ score. Patients under age 35 years or pregnant had Prospective Investigation of Pulmonary Embolism Diagnosis (PIOPED) interpreted [7] ventilation-perfusion (VQ) scanning and a CT pulmonary angiogram (CTPA) if VQ was nondiagnostic. All others had CTPA. Every patient was followed clinically for 3 months by hospital record search and telephone. Death certificates were obtained for patients who died during follow-up. An independent adjudication committee reviewed the cases where patients presented with possible VTE during follow-up, and all deaths.

The primary outcome was the area under the receiver-operating characteristics (ROC) curves for high-sensitivity troponin alone and in combination with clinical probability scoring.

Between 2nd September 2008 and 19th June 2009, 411 patients were investigated for PE in our institution. Of these, 354 consented to take part in the study, two patients could not be contacted and 55 fulfilled exclusion criteria. The mean age was 54 (± 11 years), 59% were female. In all, 71% were emergency department presentations, and the median duration of symptoms was 3 days. Twenty-two per cent had a history of ischemic heart disease and 15% were concurrently investigated for acute coronary syndrome. The mean heart rate was 88 (± 19), and 2% had a systolic BP < 90 mmHg.

PE was diagnosed in 68/354 (19%) patients. Diagnostic investigations were incomplete in 21/354 (6%). All hospital records were scrutinized after 3 months. Follow-up was completed by telephone for 259 patients and by letter for 27 patients. General practitioner notes were searched for all of the remaining 34 patients. Thirty-four patients (10%) died during follow-up.

Six patients had insufficient serum to analyse for troponin T. The median troponin level was 0.006 ng mL−1 (IQR 0.003–0.021 ng mL−1), and 31% of patients had a troponin level < 0.003 ng mL−1. The area under the ROC curve (AUC) for high-sensitivity troponin (Fig. 1) was 0.64 [95% confidence interval (CI) 0.58–0.69], with an optimal cut-off of 0.003 ng mL−1. This gives a sensitivity of 91.2% (95% CI 81.8–96.7%) and specificity of 40.2% (95% CI 34.5–46.8%). The AUC for combined high-sensitivity troponin and Wells’ score is 0.76 (95% CI 0.71–0.80). Applying a cut-off of 0.003 ng mL−1 for troponin and 4.0 for the Wells’ score (either test positive meaning a positive combined test), gives a sensitivity of 95.6% (95% CI 87.6–99.1%) and specificity of 35.1% (95% CI 29.4–41.2%). The respective AUCs for troponin alone and troponin combined with the Wells’ score in emergency department patients suspected of PE (n = 235, prevalence 20.0%) were 0.71 (95% CI 0.64–0.76) and 0.79 (95% CI 0.74–0.84), respectively.

Figure 1.

 Comparison of receiver-operating characteristics (ROC) curves. Troponin, n = 327, prevalence 20.8%, area under the curve (AUC) 0.64 [95% confidence interval (CI) 0.58–0.69]. Combined Well’s score and troponin, n = 235, prevalence 20.6%, AUC 0.76 (95% CI 0.71–0.80). Combined Well’s score and troponin in emergency department patients, n = 234, prevalence 20.1%, AUC 0.79 (95% CI 0.74–0.84). Combined Well’s score and D-dimer, n = 218, prevalence 17.4%, AUC 0.86 (95% CI 0.81–0.90).

The present study is the first to assess troponin as a potential diagnostic tool for PE. The introduction of this new, low-reading troponin T assay has allowed us to explore whether patients investigated for PE might have more marginal elevations in their absolute troponin level. High-sensitivity troponin T does not have sufficient sensitivity or specificity as a lone test; however, there was an association with PE in emergency department patients or when combined with clinical probability.

Our study resembled real-life. 1.2% of the cohort was subsequently diagnosed with ACS and 3.3% had been diagnosed with ACS in the preceding 4 weeks. Eleven percent had serum creatinine values > 120μmol L−1 and a proportion had multi-organ failure. After myocardial infarction, blood troponin levels rise to a maximum within 24 h, following which levels decline towards normal. In the present study, the median delay to investigation of symptoms was 3 days (IQR 1–10 days). Unlike cardiac chest pain, the symptoms of PE can be insidious, or masked by other co-morbidity, leading to a delay in seeking medical advice.

Our analysis has certain limitations. First, we combined high-sensitivity troponin (a blinded test) with the Wells’ score. This could have introduced bias as the Wells’ score was recorded prospectively by each clinician, and used inherently as part of the reference standard algorithm. Second, caution should be applied when comparing combined clinical probability and D-dimer. Both were part of the reference standard and therefore may appear to perform as better diagnostic tests than they are in reality. Furthermore, many patients did not have a D-dimer test because they scored moderate or high clinical probability, so the ROC curve does not represent the whole spectrum of the patient cohort. Finally, our initial sample size estimate for the THREAD study suggested that 800 patients were required to prove that a new test performed better than conventional initial assessment. The study recruited 806 patients investigated for either PE or DVT, but only 354 with suspected PE. Despite this, our confidence intervals support our summary.

To conclude, high-sensitivity troponin T cannot be used alone to diagnose PE. There is a moderate association with PE when combined with clinical probability, and in emergency department patients investigated for PE. These combinations do not perform better than conventional assessment.


A. Vylkov, B. Sethi, J. Shaw and D. Sfeir helped with patient follow-up. B. Martin, R. O’Driscoll and R. Warner formed the independent adjudication committee. The College of Emergency Medicine and Leo Pharma.

Disclosure of Conflict of Interests

The authors state that they have no conflict of interest.