Patients with chest pain constitute the largest single class of medical admissions to the Emergency Department (ED). They provide the greatest headache to the ED physician because the majority of such patients do not have the final diagnosis of acute myocardial infarction (AMI) and clog up the ED, but they also constitute the largest single group of litigation when the diagnosis of AMI is missed. The challenge has always been to have a strategy that will allow rapid exclusion of low-risk patients who can be expeditiously discharged and to identify the high-risk patients for admission. Recently, the use of structured decision protocols to identify patients at low risk of AMI have been developed and combined with serial measurement of cardiac troponin over short time periods (measurement on admission and two hours from admission). This is still a diagnosis based approach but a movement to a risk-based (prognostic) approach has been developed with the use of high sensitivity troponin assays. In this model, a single troponin measurement made on admission is combined with selection of a low risk population to use the combination of prior probability (Bayes theorem) with prognostic risk assessment from the troponin measurement. This approach does not utilise the 99th percentile (the diagnostic threshold) but a much lower value typically the limit of detection or limit of blank of the assay.
In the recent study presented at the AACC we explored how effective this approach would be utilising to contemporary sensitive (the Siemens Stratus CS and the Siemens cTnI Ultra) and two (in our hands) high sensitivity assays (the Beckman Accu I+3 prototype and the Roche hs cTnT) directly comparing these 4 assays in the same samples from a low risk chest pain population.
The samples came from one arm of a prospective randomised controlled trial which compared point of care testing (POCT) with conventional management. All the patients enrolled in the POCT arm had an additional sample taken on admission which was archived. All patients had full follow-up and outcome determined (major adverse cardiac events). We measured troponin by the four methods above and then examined diagnostic and prognostic accuracy with a discriminant based on the limit of detection of the assay.
Limit of detection measured on admission allowed accurate exclusion of MI in 98.5-99.2% of patients with a final diagnosis that excluded MI corresponding to 70.6-80.8% of all patients presenting. In those who ruled out based on a single admission measurement major adverse events occurred in 0.2-0.6% and comprised readmission with suspected acute coronary syndrome and 1 myocardial infarction on follow up. It was interesting to note the dichotomy between diagnostic accuracy and prognosis. Although diagnostic accuracy did not achieve >99.9%, the ability to predict outcome did approach this figure.
High sensitivity assays were better than contemporary sensitive assays but only just. The study illustrates that prior probability is one of the most important features in achieving a prognostically satisfactory rule out. Although high sensitivity assays will achieve a good result they will never achieve 100% diagnostic accuracy but may achieve sufficient prognostic accuracy that when used in an appropriate population will allow single test rule out.