Recently, I published a paper in the Journal of the American College of Cardiology describing a strategy that could rule out acute myocardial infarction (AMI) in over a quarter of patients using a single blood test at the point of presentation.
The strategy relies on the use of the high sensitivity troponin T assay from Roche Diagnostics. This assay has a 99th percentile (the diagnostic cut-off) of 14ng/L but the assay can detect levels way below that, in apparently healthy people. In fact, the assay can be reported down to 3ng/L, which is known as the limit of blank (LoB) of the assay.
We know that high sensitivity troponin assays can quantify levels of troponin below the 99th percentile but, at present, we don’t use that information clinically. It doesn’t matter whether the patient has completely undetectable levels of high sensitivity troponin T (<3ng/L) or levels of anything from 3 to 14ng/L. So long as the level is <14ng/L, it’s normal – and that’s all we need to know to rule out AMI.
Ever since high sensitivity assays became available, we’ve also known that they have a higher diagnostic sensitivity for AMI than previous generation troponin assays, i.e. more of the patients with AMI will have positive results when they first arrive at hospital. Unfortunately, the sensitivity still isn’t good enough to allow us to reassure patients that they’re not having a heart attack without repeating the test a number of hours later – at least, not using the conventional cut-off.
I wondered whether we could use the additional information from levels below the 99th percentile to enable us to rule out AMI in some patients immediately. It makes sense that, if troponin levels slowly begin to rise in the circulation after AMI, there should be detectable levels in the circulation even before the level exceeds the 99th percentile. Perhaps those patients who have absolutely no detectable troponin using this highly sensitive assay could be immediately reassured that they don’t have AMI.
Tobias Reichlin’s seminal publication in the New England Journal of Medicine in 2009 suggests that this strategy may have some mileage. In the online appendix, the sensitivity of high sensitivity troponin T (hs-cTnT) at the 3ng/L cut-off was 100%, i.e. no patients with initially undetectable (<3ng/L) levels had AMI.
We sought to validate that finding. In a prospective diagnostic cohort study, we had serum samples from 703 patients. Among these patients, a hs-cTnT level of <3ng/L had 100% sensitivity and negative predictive value for AMI. We then went on to audit our results after implementing hs-cTnT in the clinical environment. Among 915 patients who underwent 2 or more hs-cTnT tests during their admission at our institution (for any reason), only 1 patient with initially undetectable (<3ng/L) troponin developed a subsequent rise, giving a sensitivity of 99.4%. These findings formed the basis of the publication in the Journal of the American College of Cardiology, at http://www.ncbi.nlm.nih.gov/pubmed/21920261.
Having documented these findings, we were keen to determine whether they would stand up to prospective validation and could be introduced into practice to rule out AMI earlier and enable earlier reassurance for many patients, while reducing unnecessary hospital admissions. We therefore undertook a systematic review.
The systematic review is ongoing but early findings are to be presented at the International Conference on Emergency Medicine in June 2012. From 7 studies including 4,629 patients hs-cTnT at a cut-off of <3ng/L was found to have a sensitivity of 99.7% (95% CI 98.4 – 100.0%) and negative predictive value 99.8% (CI 98.7% – 100.0%). This would have enabled 17.5% of patients to have AMI excluded.
99.7% is a pretty good sensitivity, let’s face it. What’s more, it’s better than 98.2% sensitivity of a 0- and 3-hour high sensitivity troponin (Abbott Architect hs-troponin I), as reported recently in JAMA. In fact, not many diagnostic tests in medicine have such good sensitivity.
Of course, this systematic review is still ongoing and there is more recent evidence to synthesise. However, given the interim results, the question is whether you’d be happy to exclude AMI based on this evidence. And, if not, why not?
I’m not asking these questions to be provocative and I’m not stating a personal opinion. I’m asking these questions to stimulate discussion. I’ll be presenting the systematic review at ICEM on Saturday, 30th June just after 2pm. In the meantime, I’d like to harness the power of social media and invite discussion of these findings and their acceptability for practice, via Twitter.
Please feel free to comment, to question, to criticise. And please use the hashtags #hsTnT and #ICEM2012. My presentation at ICEM has these Twitter feeds embedded so we’ll be checking in with your Tweets, live at ICEM, at the end of the presentation. Thanks in advance for your opinions!