5 Outcomes

The Diagnostics Advisory Committee (section 11) considered evidence from a number of sources (section 12).

How outcomes were assessed

5.1 The assessment consisted of a systematic review of the evidence on test performance and clinical‑effectiveness data for the Elecsys Troponin T high‑sensitive, ARCHITECT STAT High Sensitive Troponin‑I and AccuTnI+3 assays.

Clinical effectiveness

5.2 The External Assessment Group conducted a systematic review of the evidence on the clinical effectiveness of high‑sensitivity troponin testing for early rule out or diagnosis of acute myocardial infarction (MI) within 4 hours of people presenting with acute chest pain at an emergency department. Studies were considered for inclusion based on criteria developed for each of the clinical‑effectiveness questions defined in the review protocol.

5.3 In total, 18 studies reported in 38 publications were included in the review. Of these, 15 studies (reported in 34 publications) reported accuracy data for the Elecsys Troponin T high‑sensitive assay and 4 studies (reported in 5 publications) reported accuracy data for the ARCHITECT STAT High Sensitive Troponin‑I assay. No published studies for the AccuTnI+3 assay were identified. The External Assessment Group identified 2 studies (reported in 3 publications) reporting accuracy data for an assay described as a prototype Access high‑sensitivity troponin I assay. Some studies included more than 1 high‑sensitivity troponin assay. The External Assessment group indicated that it was not clear from the studies whether this prototype was designed as a previous version of the AccuTnI+3 assay. The data for this prototype were included by the External Assessment Group in the diagnostics assessment report because the assay was developed by Beckman Coulter and its characteristics met the criteria used for defining high‑sensitivity in the scope. All 18 studies were classed as diagnostic cohort studies; no randomised controlled trials or controlled clinical trials were identified. Of the 18 identified studies, 13 were conducted in Europe (of which 2 were UK‑based), 4 were conducted in Australia and New Zealand, and 1 was conducted in the USA. In addition, data submitted to the FDA relating to the AccuTnI+3 assay were provided by Beckman Coulter. The manufacturer considered these data to be commercial in confidence at the time of publication of the guidance.

5.4 Critical appraisal of the identified studies was done using the QUADAS‑2 tool. The main potential sources of bias in the included studies related to patient spectrum and patient flow, which were of particular concern in studies that excluded people presenting out of hours or that stopped recruitment when workload was high. There were concerns about applicability of the patient population; 7 studies excluded people with STEMI and 1 study reported data separately for people in whom STEMI was excluded. The remaining studies included in the assessment reported data from a mixed population; that is, for which the target condition was either NSTEMI or STEMI (any acute MI). There were also concerns about the applicability of the reference standard due to the review criteria stating that an appropriate reference standard was standard troponin measurement at baseline and at 10–12 hours in 80% of the study population (only 5 studies met this criterion).

Evidence on diagnostic accuracy

5.5 For meta‑analyses including 4 or more studies, the External Assessment Group used the bivariate/hierarchical summary receiver operator characteristic model to estimate summary sensitivity and specificity with 95% confidence intervals and prediction regions. For meta‑analyses involving fewer than 4 studies, the External Assessment Group calculated separate pooled estimates of sensitivity and specificity using random‑effects logistic regression. Analyses were performed separately for each high‑sensitivity troponin assay, and were stratified according to whether the study reported the prediction of acute MI or major adverse cardiac events, timing of the collection of the blood sample for troponin testing, and the threshold used to derive a positive high‑sensitivity troponin result. Where possible, data were reported for all studies, studies excluding people with STEMI and studies in which the target condition was any acute MI.

5.6 The External Assessment Group presented the results of its analyses according to whether the studies reported results from samples taken at the time the patient presented to the emergency department (presentation samples), samples taken between 1 and 3 hours after the patient presented to the emergency department (subsequent samples), and diagnostic strategies that involved multiple samples (multiple samples).

Diagnostic accuracy of the Elecsys Troponin T high‑sensitive assay

5.7 Fourteen studies assessed the accuracy of the Elecsys Troponin T high‑sensitive assay for the detection of acute MI.

Presentation samples

5.8 All 14 studies reported accuracy for the detection of acute MI on single samples taken at presentation. All but one of the studies reported data on presentation samples using the 99th percentile as the diagnostic threshold, providing a summary estimate of sensitivity of 89% (95% confidence interval[CI] 85% to 92%) and specificity of 82% (95% CI 77% to 86%). The positive likelihood ratio (LR+) – how many more times likely it is that a person with the target condition will receive a positive test result compared with a person without the condition – was 4.96 (95% CI 3.84 to 6.39). The negative likelihood ratio (LR−) (how many more times likely it is that a person with the target condition will receive a negative test result compared with a person without the condition) was 0.14 (95% CI 0.10 to 0.19). When this analysis was restricted to studies that excluded people with STEMI (6 studies), similar summary estimates were obtained: sensitivity 88% (95% CI 78% to 93%), specificity 84% (95% CI 74% to 90%), LR+ 5.41 (95% CI 3.40 to 8.63) and LR− 0.15 (95% CI 0.08 to 0.26). The External Assessment Group also repeated this analysis when restricted to studies with a mixed population, that is when the target condition was any acute MI (8 studies), and similar summary estimates were obtained: sensitivity 89% (95% CI 86% to 91%), specificity 81% (95% CI 76% to 85%), LR+ 4.64 (95% CI 3.73 to 5.76) and LR− 0.14 (95% CI 0.11 to 0.17). The External Assessment Group considered that the analysis which excluded people with STEMI was the most applicable to the population defined in the scope and this was used to inform the cost‑effectiveness analysis.

5.9 Data on presentation samples using diagnostic thresholds equivalent to the limit of detection (5 nanograms/litre) or limit of blank (3 nanograms/litre) were reported in 5 studies. Three studies reported data for the limit of detection, 1 of which was excluded from analyses because the limit of detection data were only reported for people over 70 years of age. The summary estimates for presentation samples at the limit of detection were sensitivity 95% (95% CI 92% to 97%), specificity 54% (95% CI 51% to 58%), LR+ 2.06 (95% CI 1.40 to 2.64) and LR− 0.09 (95% CI 0.07 to 0.17). One study reported these data excluding people with STEMI: sensitivity 93% (95% CI 89% to 96%), specificity 58% (95% CI 55% to 62%), LR+ 2.20 (95% CI 2.00 to 2.50) and LR− 0.11 (95% CI 0.07 to 0.19). Three studies reported data for the limit of blank: summary estimates were sensitivity 98% (95% CI 95% to 99%), specificity 40% (95% CI 38% to 43%), LR+ 1.63 (95% CI 1.24 to 1.86) and LR− 0.05 (95% CI 0.02 to 0.21). One study reported the limit of blank in a population that excluded people with STEMI and was used to inform the cost‑effectiveness modelling; the results were sensitivity 95% (95% CI 92% to 98%), specificity 48% (95% CI 44% to 51%), LR+ 1.83 (95% CI 1.70 to 1.97) and LR− 0.10 (95% CI 0.05 to 0.18).

5.10 The External Assessment Group identified limited data on clinically relevant subgroups, including people aged less than 70 years compared with those aged 70 years or over, no pre‑existing coronary artery disease compared with pre‑existing coronary artery disease, and high pre‑test probability compared with low‑to‑moderate pre‑test probability. All of the studies that reported subgroup data were in a mixed population, and did not exclude people with STEMI. The External Assessment Group concluded that the results of subgroup analyses suggest that high‑sensitivity troponin testing, using the 99th percentile diagnostic threshold on a sample taken at presentation, could be adequate for rule out of acute MI in people aged 70 years or older (LR− 0.05 for people aged 70 years or older compared with LR− 0.14 for people aged less than 70 years), people who do not have pre‑existing coronary artery disease (LR− 0.07 for people without pre‑existing coronary artery disease compared with LR− 0.12 for people with pre‑existing coronary artery disease) and people who are classified as having a high pre‑test probability of acute MI (LR− 0.09 for high pre‑test probability compared with LR− 0.13 for low‑to‑moderate pre‑test probability).

5.11 The External Assessment Group noted that the time between onset of chest pain and presentation at the emergency department was inconsistently reported in the included studies. In studies that did report time from chest pain onset, the median time ranged from 2.7 to 8.25 hours. Two studies, conducted in a mixed population, stratified patients as presenting either before or after 3 hours since the onset of chest pain, and in addition, either before or after 6 hours since the onset of chest pain. The results of this analysis suggested that the Elecsys Troponin T high‑sensitive assay, with the 99th percentile as the diagnostic threshold on a sample taken at presentation, had a higher sensitivity (94% [95% CI 92% to 96%]) and lower specificity (77% [95% CI 75% to 79%]) for any acute MI in people presenting more than 3 hours after the onset of chest pain compared with people presenting within 3 hours; sensitivity 78% (95% CI 71% to 83%) and specificity 84% (95% CI 81% to 86%). When the analysis was repeated using 6 hours as the threshold, the results were similar.

Subsequent samples

5.12 Of the 14 studies reporting accuracy of the Elecsys Troponin T high‑sensitive assay for the detection of acute MI, 2 reported data on samples taken 1 to 3 hours after presentation. Both studies used the 99th percentile as the diagnostic threshold and excluded people with STEMI. Summary estimates were sensitivity 95% (95% CI 92% to 97%), specificity 80% (95% CI 77% to 82%), LR+ 4.75 (95% CI 3.98 to 5.23) and LR− 0.06 (95% CI 0.00 to 0.63).

Multiple samples

5.13 Of the 14 studies reporting accuracy of the Elecsys Troponin T high‑sensitive assay for the detection of acute MI, 6 reported data on the performance of diagnostic strategies involving multiple samples. The most commonly reported strategies were those involving a combination of a peak troponin T value above the 99th percentile diagnostic threshold, and a 20% change in high‑sensitivity troponin T over 2 to 3 hours after presentation; 1 study reported these data in a population that excluded people with STEMI, and consequently was used in the cost‑effectiveness modelling. The results of this analysis suggested that a test strategy defining a positive result as a peak value above the 99th percentile diagnostic threshold and a delta change in troponin T levels (that is a change from baseline) of greater than 20% over 2 hours provided the optimal rule‑in performance (LR+ of 8.42, 95% CI 6.11 to 11.60). Conversely, a test strategy defining a negative result as no value above the 99th percentile diagnostic threshold and a delta change of less than 20% over 2 hours provided the optimal rule‑out performance (LR− 0.04, 95% CI 0.02 to 0.10).

Elecsys Troponin T high‑sensitive assay test pathways

5.14 The External Assessment Group constructed an optimal test pathway for multiple sampling on which the cost‑effectiveness modelling was based. This test pathway was based on the optimal diagnostic accuracy reported in the External Assessment Group's clinical effectiveness analysis, and included data only from studies that excluded people with STEMI. For the Elecsys Troponin T high‑sensitive assay, this test pathway comprised a rule‑out step of a presentation sample with the limit of blank (3 nanograms/litre) as the diagnostic threshold. This was followed by a second sample at 2 hours with the 99th percentile threshold and delta change of more or less than 20% as the diagnostic threshold, in which a peak value less than the 99th percentile, in combination with a delta change of less than 20%, was used as a second rule‑out step. In addition, the results could also be combined when a peak value over the 99th percentile and a delta change of greater than 20% may have provided optimum rule‑in performance. Combining the results of tests over 2 hours could also create a third category, in addition to the optimal rule‑in and rule‑out categories, which included people with either a peak value above the 99th percentile diagnostic threshold or a delta change of greater than 20% who needed further investigations to determine whether they were experiencing an acute MI. To construct the test pathway used in the economic modelling, the External Assessment Group assumed that the diagnostic performance of the second sample taken at 2 hours was the same for people in whom NSTEMI was not ruled out by the presentation sample reported at the limit of blank, as for the initial population presenting to the emergency department.

5.15 When the External Assessment Group applied this test pathway to a hypothetical cohort of 1000 people presenting with a suspected non‑ST‑segment elevation acute coronary syndrome, with an estimated prevalence of NSTEMI of 17%, the first step of the strategy could result in the discharge of 407 (40.7%) people, 9 (0.9%) of whom would have been discharged in error. The second stage of the strategy could lead to the discharge of a further 286 (28.6%) people, 5 (0.5%) of whom would have been discharged in error. The External Assessment Group also applied the hypothetical cohort to the Elecsys Troponin T high‑sensitive 99th percentile presentation sample test strategy, and the number of people potentially discharged in error was 20 (2%).

Diagnostic accuracy of the ARCHITECT STAT High Sensitive Troponin‑I assay

5.16 Of the 4 diagnostic cohort studies reporting data on the ARCHITECT STAT High Sensitive Troponin‑I assay, 3 assessed the accuracy of the assay for the detection of acute MI. All 3 studies were conducted in a mixed population (any acute MI), and did not exclude people with STEMI.

Presentation samples

5.17 All 3 studies reporting accuracy of the ARCHITECT STAT High Sensitive Troponin‑I assay for the detection of acute MI reported data for single samples taken at presentation. Summary estimates based on the 99th percentile as the diagnostic threshold were sensitivity 80% (95% CI 77% to 83%), specificity 93% (95% CI 92% to 94%), LR+ 11.47 (95% CI 9.04 to 16.19) and LR− 0.22 (95% CI 0.16 to 0.27).

5.18 One study also reported the accuracy of the ARCHITECT STAT High Sensitive Troponin‑I assay based on the limit of detection as the diagnostic threshold. The results were sensitivity 100% (95% CI 98% to 100%), specificity 35% (95% CI 32% to 38%), LR+ 1.54 (95% CI 1.47 to 1.62) and LR− 0.01 (95% CI 0.00 to 0.08).

Subsequent samples

5.19 Of the 3 studies reporting accuracy of the of the ARCHITECT STAT High Sensitive Troponin‑I assay for the detection of acute MI, 1 reported data on samples taken 3 hours after presentation. The study reported accuracy based on the 99th percentile as the diagnostic threshold; the results were sensitivity 98% (95% CI 96% to 99%), specificity 90% (95% CI 88% to 92%), LR+ 10.16 (95% CI 8.38 to 12.31), LR− 0.02 (95% CI 0.01 to 0.08).

Multiple samples

5.20 Of the 3 studies reporting accuracy of the ARCHITECT STAT High Sensitive Troponin‑I assay for the detection of acute MI, 2 reported data on the performance of diagnostic strategies involving multiple samples. One study reported the accuracy of a sample taken at presentation and at 2 to 3 hours with a peak troponin I value over the 99th percentile. The second study reported the accuracy of a sample taken on admission with the limit of detection as the threshold and a delta change of 20% at 3 hours, and the accuracy of a sample taken on presentation, and at 3 hours, with a delta change of 20%. The results of these studies suggested that the sensitivity and specificity of multiple sampling strategies could range from 77% to 91% and from 26% to 93% respectively, depending on the timing of sampling and the diagnostic threshold applied.

ARCHITECT STAT High Sensitive Troponin‑I assay test pathways

5.21 The External Assessment Group constructed an optimal test pathway for multiple sampling on which the cost‑effectiveness modelling was based. This test pathway was derived from the results of the External Assessment Group's clinical‑effectiveness review, and included estimates taken from studies reporting diagnostic accuracy in a mixed population. For the ARCHITECT STAT High Sensitive Troponin‑I assay this test pathway comprised a presentation sample with the limit of detection as the diagnostic threshold as an initial rule‑out step, followed by a second sample at 3 hours with the 99th percentile as the diagnostic threshold. To construct the test pathway used in the economic modelling the External Assessment Group assumed that the diagnostic performance of the second sample taken at 3 hours was the same for people in whom NSTEMI was not ruled out by the presentation sample reported at the limit of detection, as for the initial population presenting to the emergency department.

5.22 The External Assessment Group applied this test pathway to a hypothetical cohort of 1000 people presenting with a suspected non‑ST‑segment elevation acute coronary syndrome and an estimated prevalence of NSTEMI of 17%. The first step of the strategy could result in the discharge of 291 (29.1%) people, none of whom would have been discharged in error. The second stage of the strategy could lead to the discharge of a further 486 (48.6%) people, 3 (0.3%) of whom would have been discharged in error. This cohort was also applied to the ARCHITECT STAT High Sensitive Troponin‑I 99th percentile presentation test strategy and led to 34 (3.4%) people potentially being discharged in error.

Diagnostic accuracy of the AccuTnI+3 troponin I assay

5.23 Both of the diagnostic cohort studies reporting data on the prototype Access high‑sensitivity troponin I assay assessed the accuracy of the assay for the detection of acute MI. Both studies were conducted in a mixed population (any acute MI), and did not exclude people with STEMI. It should be noted that each of these studies report the performance of a prototype assay, with test characteristics that differ from those provided by the manufacturer for the AccuTnI+3 assay. These data were included by the External Assessment Group in their report because this prototype assay was developed by Beckman Coulter and the assay's characteristics met the criteria used for defining high‑sensitivity in the scope and, therefore, it may be considered relevant by the Committee.

Presentation samples

5.24 Both studies reporting the accuracy of the prototype Access high‑sensitivity troponin I assay for the detection of acute MI reported data on single samples taken at presentation. One study reported the diagnostic accuracy of a sample taken on presentation using the 99th percentile as the diagnostic threshold (note that this was 9 nanograms/litre for the assay used in this study, described as an investigational prototype). The results of the study were sensitivity 92% (95% CI 88% to 95%), specificity 75% (95% CI 72% to 78%), LR+ 3.67 (95% CI 3.26 to 4.13) and LR− 0.11 (95% CI 0.07 to 0.17). When the results of this study were combined with the second study, which reported presentation samples at a diagnostic threshold of 18 nanograms/litre, the summary estimates of accuracy were sensitivity 92% (95% CI 88% to 95%), specificity 75% (95% CI 72% to 77%), LR+ 3.68 (95% CI 2.46 to 4.48) and LR− 0.11 (95% CI 0.07 to 0.16).

Subsequent samples

5.25 None of the studies reporting data on the prototype Access high‑sensitivity troponin I assay provided information on subsequent samples.

Multiple samples

5.26 Of the 2 studies reporting accuracy for the detection of acute MI, 1 reported data on the diagnostic performance of a 27% or greater change in troponin I levels between presentation and 1 hour. The results were sensitivity 63% (95% CI 53% to 71%), specificity 66% (95% CI 63% to 69%), LR+ 1.85 (95% CI 1.55 to 2.21) and LR− 0.56 (95% CI 0.44 to 0.72).

AccuTnI+3 troponin I test pathways

5.27 For the AccuTnI+3 assay economic modelling, the External Assessment Group included a single test strategy which comprised a presentation sample using the prototype Access high‑sensitivity troponin I assay with the 99th percentile (9 nanograms/litre) as the diagnostic threshold, and was based on a study that did not exclude people with STEMI. When this test pathway was applied to a hypothetical cohort of 1000 people presenting with a suspected non‑ST‑segment elevation acute coronary syndrome and an estimated prevalence of NSTEMI of 17%, the number of people with acute MI potentially discharged in error on the basis of a negative high‑sensitivity troponin I result was 14 (1.4%). The External Assessment Group did not construct a test pathway with a second sample because of the limited data available on this assay.

Comparative accuracy of the high‑sensitivity troponin assays

5.28 The External Assessment Group derived summary estimates from analyses with common time points and cut‑off thresholds to compare the accuracy of the 3 high‑sensitivity troponin assays. In addition, 1 study provided a direct comparison of all 3 assays in the same population. This study reported data on presentation samples using the 99th percentile as the diagnostic threshold. The study reported a sensitivity of 90% for the Elecsys Troponin T high‑sensitive assay, 77% for the ARCHITECT STAT High Sensitive Troponin‑I assay and 92% for the Access high‑sensitivity troponin I assay. Corresponding values for specificity were 78%, 93% and 75% for each assay respectively. Summary estimates derived from the indirect analysis were similar.

Evidence on prognostic accuracy

Prognostic accuracy of the Elecsys Troponin T high‑sensitive assay

5.29 Of the 15 diagnostic cohort studies reporting data on the Elecsys Troponin T high‑sensitive assay, 1 assessed the accuracy of the assay for the prediction of major adverse cardiac events within 30 days of presentation. The final scope for this assessment defined major adverse cardiac events as death, non‑fatal acute MI, revascularisation or hospitalisation for myocardial ischaemia. The study, which excluded people with STEMI, reported the prognostic accuracy of a presentation sample using the limit of blank (3 nanograms/litre) as the threshold. The study reported a sensitivity of 85% (95% CI 74% to 92%), specificity 46% (95% CI 41% to 51%), LR+ 1.58 (95% CI 1.37 to 1.81) and LR− 0.33 (95% CI 0.18 to 0.59).

Prognostic accuracy of the ARCHITECT STAT High Sensitive Troponin‑I assay

5.30 Of the 3 diagnostic cohort studies reporting data on the ARCHITECT STAT High Sensitive Troponin‑I assay, 1 assessed the accuracy of the assay for the prediction of major adverse cardiac events within 30 days of presentation. The study, in a mixed population (any acute MI), reported the prognostic accuracy of a presentation sample using the 99th percentile as the threshold. The study reported a sensitivity of 88% (95% CI 85% to 91%), specificity 93% (95% CI 91% to 94%), LR+ 12.57 (95% CI 8.88 to 15.35) and LR− 0.13 (95% CI 0.06 to 0.28).

Prognostic accuracy of the AccuTnI+3 troponin I assay

5.31 Neither of the 2 diagnostic cohort studies reporting data on the prototype Access high‑sensitivity troponin I assay assessed the accuracy of the assay for the prediction of major adverse cardiac events within 30 days of presentation.

Costs and cost effectiveness

5.32 The External Assessment Group conducted a systematic review to identify existing studies investigating the cost effectiveness of diagnostic strategies for acute coronary syndrome which incorporated high‑sensitivity troponin testing. Studies reporting a full economic analysis relating to the cost‑effectiveness of either high‑sensitivity troponin or standard troponin testing, which included survival or quality‑adjusted life years (QALYs) as an outcome measure were eligible for inclusion.

5.33 The External Assessment Group also constructed a de novo economic model designed to assess the cost effectiveness of the Elecsys Troponin T high‑sensitive assay, the ARCHITECT STAT High Sensitive Troponin‑I assay and the AccuTnI+3 troponin I assay used singly or in series up to 4 hours from the onset of chest pain or presentation to an emergency department.

Systematic review of cost effectiveness

5.34 The systematic review identified 5 studies, reported in 7 publications. The included studies evaluated a range of diagnostic strategies for acute MI, including the use of both high‑sensitivity troponin and standard troponin testing. The diagnostic strategies included using high‑sensitivity troponin testing alone, combining high‑sensitivity troponin testing with heart‑fatty acid binding protein, panels of cardiac biomarkers and point of care cardiac biomarker panel testing. The timing of tests varied both within and between the included studies.

5.35 The results of the studies included in the systematic review varied widely, and the External Assessment Group concluded that the review demonstrated uncertainty about the cost effectiveness of diagnostic strategies incorporating high‑sensitivity troponin testing. The External Assessment Group noted that the key drivers of cost effectiveness in the included studies were the accuracy of high‑sensitivity troponin assays, and the efficiency of decision‑making once test results were available.

Economic analysis

5.36 The External Assessment Group developed a de novo economic model designed to assess the cost effectiveness of 5 high‑sensitivity troponin test strategies using 3 high‑sensitivity assays: the Elecsys Troponin T high‑sensitive assay, the ARCHITECT STAT High Sensitive Troponin‑I assay and the AccuTnI+3 troponin I assay. The population included in the economic model was people presenting to the emergency department with a suspected non‑ST‑segment elevation acute coronary syndrome (with STEMI ruled out), who had no major comorbidities needing hospital admission, such as heart failure or arrhythmia. Expected costs, life years and QALYs were calculated for each of the diagnostic strategies included in the model; discount rates of 3.5% and a half‑cycle correction were applied for both costs and effects.

5.37 The following high‑sensitivity troponin test strategies were included in the model:

  • Single test strategies

    • Elecsys Troponin T high‑sensitive assay presentation sample with the 99th percentile as the diagnostic threshold

    • ARCHITECT STAT High Sensitive Troponin‑I assay presentation sample with the 99th percentile as the diagnostic threshold

    • AccuTnI+3 troponin I presentation sample (using data from the prototype Access high‑sensitivity troponin I assay with the 99th percentile [9 nanograms/litre] as the diagnostic threshold).

  • Sequential test strategies

    • Elecsys Troponin T high‑sensitive assay optimal test strategy (see section 5.14)

    • ARCHITECT STAT High Sensitive Troponin‑I assay optimal test strategy (see section 5.21)

Model structure

5.38 The model structure was based upon a previous cost‑effectiveness model reported in Goodacre et al. (2013). The model was adapted to fit the scope of this assessment.

5.39 The model comprised a decision tree to model the short‑term (30 days) outcomes after presentation and a long‑term Markov model. The decision tree modelled the results of high‑sensitivity troponin or standard troponin testing and the resulting treatment decision, be it hospital admission and acute MI treatment or discharge from the emergency department. The outcomes included in the short‑term model were 'no acute coronary syndrome, no unstable angina', 'unstable angina', 'untreated non‑fatal acute MI', 'treated non‑fatal acute MI' and 'death'. The long‑term outcomes, based on a 60‑year lifetime time horizon, were estimated using a Markov cohort model with a 1‑year cycle time. The following health states were included in the model: 'no acute coronary syndrome, no unstable angina', 'unstable angina', 'post‑acute MI (treated and untreated)', 'post‑acute MI with re‑infarction' and 'death'. People entered the Markov model based on their short‑term outcome from the decision tree; consequently, in the first cycle, the model made a distinction between treated and untreated acute MI.

Model inputs

5.40 The model was populated using data from the published literature including previous economic evaluations and literature retrieved to inform key parameters (such as acute MI prevalence), routine sources of cost data and, when necessary, through consultation with experts for unpublished data. The test accuracy estimates used in the model were derived from the External Assessment Group's clinical effectiveness review. The diagnostic strategies included in the model were selected based upon optimal diagnostic performance derived from the External Assessment Group's clinical effectiveness review and, when possible, data were restricted to studies that excluded people with STEMI. Estimates of test accuracy for the ARCHITECT STAT High Sensitive Troponin‑I and prototype Access high‑sensitivity troponin I assays were based on studies that did not exclude people with STEMI.

Costs

5.41 Data on costs and resource use associated with the diagnostic strategies were estimated using previous economic evaluations, routinely available data on resource use and event costs, information provided by manufacturers and, when necessary, input from experts. Estimates of test‑specific resource use were informed by the results of the External Assessment Group's clinical effectiveness review. It was assumed that acute MI treatment may include aspirin, statins, and angiotensin‑converting enzyme inhibitors with consideration of coronary revascularisation for people considered as being at high‑risk of major adverse cardiac events. It was also assumed that starting acute MI treatment for people with NSTEMI would reduce the probability of major adverse cardiac events, including cardiac death and re‑infarction. The model also assumed that the average cost of a troponin test (high‑sensitivity or standard) to the NHS is £20 (Goodacre et al, 2013). This estimate includes the cost of the assay reagents, the cost of the analyser and maintenance, and costs associated with calibration and quality control.

Health‑state utilities

5.42 Health‑state utility scores were obtained from the published literature. Age‑dependent utility scores, based on the UK general population, were calculated for people in the 'no acute coronary syndrome, no unstable angina' health state based on a linear regression model. The disutility for age was estimated to be 0.004. To calculate utility scores for the 'post‑MI' health states, age‑dependent utility scores from the general population were combined with age‑dependent disutilities for acute MI. Utility scores for the 'unstable angina' health state were then based upon the 'post‑MI' utilities, with a utility increment of 0.010.

Base‑case analysis

5.43 The assumptions applied in the base‑case analysis included the following:

  • The comparator, serial troponin testing over 10–12 hours, has perfect diagnostic accuracy (sensitivity 1.0 and specificity 1.0).

  • For the Elecsys Troponin T high‑sensitive assay and ARCHITECT STAT High Sensitive Troponin‑I optimal testing strategies, the sensitivity and specificity for the subpopulation not discharged after the presentation test is equal to the sensitivity and specificity for the initial group presenting at the emergency department.

  • The life expectancy, quality of life and costs for people with false‑positive results (that is those who are positive by the high‑sensitivity troponin test strategy but negative by standard troponin testing) is equal to the life expectancy, quality of life and costs of people with true‑negative results (this assumption was amended in the secondary analysis).

  • In contrast with acute MIs occurring in the decision tree period, all acute MIs occurring in the Markov trace are correctly diagnosed and subsequently treated.

  • Unstable angina is always correctly diagnosed and subsequently treated.

  • The re‑infarction probability for the 'post‑MI with re‑infarction' health state is equal to the re‑infarction probability for the 'post‑MI' health state.

  • The increased post‑MI re‑infarction and mortality probabilities for untreated acute MI were assumed to last 1 year; after 1 year a relative risk of 1.0 was applied (for untreated compared with treated acute MI).

  • There is no additional benefit of starting treatment early, so treatment effect for high‑sensitivity troponin testing strategies is equal to treatment effect for standard troponin testing strategies.

  • All deaths within 30 days of presentation at the emergency department are caused by fatal acute MI events and receive the associated costs.

  • Doctors are available on demand to make management decisions based on test results.

  • A total delay of 3 hours is assumed, which includes a delay between the patient presenting and the blood sample being taken, and a second delay between the sample being taken and the results becoming available.

  • No additional treatment costs are applied to people with unstable angina in year 1.

5.44 Probabilistic results were presented for the base case (based on 10,000 simulations) and were used to construct cost‑effectiveness acceptability curves and cost‑effectiveness acceptability frontiers.

Base‑case results

5.45 The base‑case analysis included 6 test strategies: 5 high‑sensitivity troponin test strategies and the comparator, standard troponin testing over 10–12 hours. The results of the base‑case analysis suggested that standard troponin testing was the most effective (15.101 life years, 11.730 QALYs) and most expensive (£2697) test strategy. In contrast, the ARCHITECT STAT High Sensitive Troponin‑I 99th percentile presentation sample was the least effective (15.076 life years, 11.712 QALYs) and least expensive (£2,253). The incremental cost‑effectiveness ratios (ICERs) for the high‑sensitivity troponin test strategies ranged from £24,019 to £90,725 saved per QALY lost compared with standard troponin.

Base‑case deterministic sensitivity analyses

5.46 The External Assessment Group also presented the results of a deterministic base‑case analysis. In this analysis, the ICERs for the high‑sensitivity troponin testing ranged from £28,870 to £124,391 saved per QALY lost. The External Assessment Group performed a number of one‑way sensitivity analyses to assess the impact of both model assumptions and input parameters on the estimated ICERs. The External Assessment Group concluded that the results of the sensitivity analyses showed that, in general, there were no major changes to relative cost effectiveness. When it was assumed that the increased re‑infarction and mortality risk for treated versus untreated acute MI lasted for a lifetime, rather than just the first year, all high‑sensitivity troponin testing strategies had ICERs under £30,000 saved per QALY lost when compared with standard troponin. In all analyses where the waiting time for a doctor on a general ward is added (1–3 hours), the test cost was increased to £40, or the acute MI treatment cost was assumed to be £4295, all high‑sensitivity troponin testing strategies had ICERs over £30,000 saved per QALY lost when compared with standard troponin. The assumption which had the most noticeable impact on the ICERs was applying acute MI treatment costs to people who received a false positive test result.

5.47 The External Assessment Group noted that input parameters which had a noticeable impact on the ICERs were as follows:

  • 30‑day mortality for treated acute MI – ICERs for high‑sensitivity troponin testing compared with standard troponin testing ranged from £41,819 to £182,781 saved per QALY lost when the 30‑day mortality was assumed to be 0.120, and from £22,206 to £94,345 saved per QALY lost when the 30‑day mortality was assumed to be 0.074.

  • 30‑day mortality for untreated acute MI – ICERs for high‑sensitivity troponin testing compared with standard troponin testing ranged from £11,153 to £45,686 saved per QALY lost when mortality was assumed to be 0.240, and when mortality was assumed to be 0.000 all 5 high‑sensitivity troponin test strategies dominated standard troponin (that is standard troponin is more expensive and less effective), although the External Assessment Group noted that a scenario where the 30‑day mortality is worse for treated than for untreated acute MI is unlikely.

  • Relative risk for mortality for treated versus untreated acute MI – ICERs for high‑sensitivity troponin testing compared with standard troponin testing ranged from £11,771 to £48,054 saved per QALY lost when the relative risk was assumed to be 3.908, and from £128,875 to £570,869 saved per QALY lost when the relative risk was assumed to be 0.901.

Base‑case subgroup analyses

5.48 The External Assessment Group performed the following subgroup analyses: sex (stratified by age), history of previous STEMI, and acute MI prevalence based on clinically relevant subgroups defined in the scope. With the exception of acute MI prevalence, the subgroups were defined on the basis that these characteristics may be associated with physiological differences in peak troponin levels. No data were found for ethnicity and people with renal disease. The subgroup analysis for MI prevalence varied from 1% to 30% and included a no‑testing strategy as a comparator to reflect the assumption that troponin testing may not be needed when clinical judgement determines that the pre‑test probability of acute MI is low.

5.49 For women, the ICERs increase with age, and by age 75 years, ICERs for all the high‑sensitivity troponin testing strategies compared with standard troponin testing were over £30,000 saved per QALY lost, with the ARCHITECT STAT High Sensitive Troponin‑I 99th percentile presentation test strategy having the lowest ICER (£32,776). The same effect of age was seen for men. However, by age 55 years, ICERs for all the high‑sensitivity troponin testing strategies compared with standard troponin testing were over £30,000 saved per QALY lost, with the ARCHITECT STAT High Sensitive Troponin‑I 99th percentile presentation test strategy having the lowest ICER (£30,338). The External Assessment Group also conducted a subgroup analysis of acute MI prevalence which suggested that when acute MI prevalence is 1% (compared with 17% in the base‑case analysis), the no‑testing strategy had an ICER of £96,456 saved per QALY lost. These subgroup analyses were performed with non‑subgroup‑specific accuracy data, and on this basis it is likely that there is substantial uncertainty surrounding the cost effectiveness of the high‑sensitivity troponin testing strategies in the reported subgroups.

5.50 Subgroup analyses were also undertaken based on test accuracy and acute MI prevalence derived from the External Assessment Group's clinical effectiveness review. The following subgroups were considered in these analyses: aged 70 years or under and aged over 70 years, people with and without pre‑existing coronary artery disease and symptom onset less than 3 hours before presentation or more than 3 hours before presentation. These analyses could only be performed for the Elecsys Troponin T high‑sensitive 99th percentile presentation strategy because of the availability of subgroup data in the clinical‑effectiveness review. In all subgroups the Elecsys Troponin T high‑sensitive assay was less costly and less effective than standard troponin testing, but greater cost savings were reported for people aged younger than 70 years, people with pre‑existing coronary artery disease, and people presenting within 3 hours of symptom onset. The limited availability of subgroup specific accuracy data means that it is likely that there is substantial uncertainty surrounding the cost effectiveness of the high‑sensitivity troponin testing strategies in the reported subgroups.

Secondary analysis

5.51 The External Assessment Group also conducted a secondary analysis, which adjusted the risk of re‑infarction and mortality for people with a false‑positive test result (that is, those with a positive high‑sensitivity troponin test result but negative standard troponin test result). This assumption is based on data that suggest that people who have a false‑positive high‑sensitivity troponin test result have an increased risk of re‑infarction and mortality compared with people with a true negative test result. The secondary analysis assumed that the prevalence of this higher‑risk subgroup was equal to the lowest proportion of people with false‑positive results for all high‑sensitivity troponin test strategies. The higher‑risk subgroup was assumed to be treated in all high‑sensitivity troponin test strategies, accruing the same relative treatment benefit as people with a true positive test result, but remained untreated for the standard troponin test. The post‑MI utility and health state costs were also applied to the higher‑risk subgroup.

5.52 For the secondary analysis, probabilistic results based on 10,000 simulations were presented, and used to construct cost‑effectiveness acceptability curves and cost‑effectiveness acceptability frontiers.

Secondary analysis results

5.53 The secondary analysis included the same test strategies as those reported in the base‑case analysis. The results of the secondary analysis suggested that standard troponin testing was the least effective (14.785 life years, 11.464 QALYs) and most expensive (£3058) test strategy, and was dominated by all 5 high‑sensitivity troponin test strategies. The ARCHITECT STAT High Sensitive Troponin‑I 99th percentile threshold presentation sample test strategy was the least effective of the 5 high‑sensitivity troponin testing strategies (14.833 life years, 11.501 QALYs), but was the overall least expensive test strategy (£2781). The ARCHITECT STAT High Sensitive Troponin‑I optimal test strategy was the most effective (14.855 life years, 11.518 QALYs) test strategy, and also the most expensive (£3018).

Secondary analysis deterministic sensitivity analyses

5.54 The External Assessment Group also conducted a deterministic secondary analysis. As with the probabilistic secondary analysis, all high‑sensitivity troponin test strategies dominated standard troponin testing. The External Assessment Group did a number of one‑way sensitivity analyses to assess the impact of both model assumptions and input parameters on the estimated ICERs. In most sensitivity analyses, the 5 high‑sensitivity troponin testing strategies dominated standard troponin testing. The External Assessment Group concluded that the results of the sensitivity analyses showed that, in general, there were no major changes to relative cost effectiveness.

5.55 When the waiting time to see a doctor was increased to 3 hours, the ARCHITECT STAT High Sensitive Troponin‑I optimum strategy no longer dominated standard troponin, with an ICER of £390 per QALY gained compared with standard troponin testing. This assumption did not affect the dominance of the other high‑sensitivity troponin test strategies. When acute MI treatment costs were added for people with false‑positive results, the 2 ARCHITECT STAT High Sensitive Troponin‑I test strategies dominated standard troponin testing, Elecsys Troponin T high‑sensitive 99th percentile presentation strategy had an ICER of £2065 per QALY gained, the Elecsys Troponin T high‑sensitive optimal strategy had an ICER of £6360 per QALY gained and the AccuTnI+3 troponin I 99th percentile presentation strategy had an ICER of £9142 per QALY gained compared with standard troponin testing.

5.56 The External Assessment Group noted that the following input parameters had a noticeable impact on the estimated cost effectiveness of the secondary analysis: increased test cost (£40 per test), 30‑day mortality for both treated and untreated acute MI in the decision tree, and the relative risk of mortality for treated compared with untreated acute MI in the Markov trace, although high‑sensitivity troponin testing did remain dominant compared with standard troponin testing in each of these sensitivity analyses.

Secondary analysis subgroup analyses

5.57 The subgroups included in the secondary analysis were the same as those included in the base‑case subgroup analyses. In both the sex and age subgroup analyses, high‑sensitivity troponin testing dominated standard troponin testing. However, the External Assessment Group noted that ICERs were slightly higher for men and appeared to increase with age. There did not appear to be a substantial difference in cost effectiveness between acute MI prevalence subgroups. However, at a 1% prevalence of acute MI, the ICER for the no‑testing strategy was £96,456 saved per QALY lost when compared with standard troponin testing. These subgroup analyses were done with non‑subgroup specific accuracy data and, on this basis, it is likely that there is substantial uncertainty surrounding the cost effectiveness of the high‑sensitivity troponin testing strategies in the reported subgroups.

5.58 Subgroup analyses were also done based on test accuracy and acute MI prevalence for the Elecsys Troponin T high‑sensitive 99th percentile presentation strategy derived from the External Assessment Group's clinical‑effectiveness review. The subgroups included in the secondary analysis were the same as those included in the corresponding base‑case subgroup analyses. In all subgroups, the Elecsys Troponin T high‑sensitive assay dominated standard troponin testing, but higher savings per QALY gained were reported for people aged 70 years or younger, people with pre‑existing coronary artery disease, and people presenting within 3 hours of symptom onset. The limited availability of subgroup‑specific accuracy data makes it likely that there is substantial uncertainty surrounding the cost effectiveness of the high‑sensitivity troponin testing strategies in the reported subgroups.

  • National Institute for Health and Care Excellence (NICE)