The theory and practice of imaging outcomes research

2334 www.thelancet.com Vol 385 June 13, 2015 BS IP V EM /S cie nc e Ph ot o Li br ar y About a decade ago, triggered by unexplained growth and variability in cardiovascular testing, eff orts to redefi ne imaging quality led to an appreciation of the importance of selecting the right test for the right patient (ie, giving rise to appropriate use criteria) and measuring the eff ect of testing on outcomes, in addition to measures such as image acquisition and interpretation. Although appropriate use criteria are now an established part of evidence-based cardiovascular care, imaging outcomes have not been defi ned in a more meaningful way, partly because of the many steps needed to translate test data into knowledge (ie, diagnostic thinking), action (ie, treatments), and ultimately outcomes. Nevertheless, there have been strong calls to explore the downstream consequences of imaging as a necessary step in more fully understanding its risks and benefi ts, which would then help to establish scenarios in which imaging does, and does not, add value. This task has been made more diffi cult in the era following the COURAGE and FAME trials, in which the diagnosis of anatomical coronary artery disease has been uncoupled from a need for revascularisation, and in the post-CONFIRM era, in which non-obstructive coronary artery disease is recognised to be a powerful predictor of future events. A few trials have been done to rise to this challenge in the primary prevention area and in the triage of acute chest pain, variously using risk factor burden, clinical events, and effi ciency as endpoints. Randomised trial data assessing the eff ect of testing in the very large stable chest pain population have been scarce, but are now present in the form of two new large studies— the Computed Tomography Coronary Angiography in Patients with Suspected Angina due to Coronary Heart Disease: the Scottish Computed Tomography of the HEART (SCOT-HEART) trial reported by David Newby and colleagues in The Lancet, and the PROspective Multicenter Imaging Study for Evaluation of Chest Pain (PROMISE) trial. In SCOT-HEART, 4146 patients completing an assessment for stable chest pain were randomly assigned to CT coronary angiography (CTCA) in addition to usual care or to usual care alone, which generally included an exercise treadmill test. The trial’s primary endpoint, certainty of the attribution of symptoms to angina due to presence of coronary artery disease, showed an increase in the CTCA group (relative risk 1·79, 95% CI 1·62–1·96), as did the closely related secondary endpoint of certainty of diagnosis of coronary artery disease (2·56, 2·33–2·79). Additional secondary endpoints included diagnostic reclassifi cation and changes in diagnostic and therapeutic approaches, which all happened more frequently in the CTCA group. Although investigators recorded no diff erences in the frequency or stability of angina, rates of revascularisation, or admittance to hospital, there was a non-signifi cant reduction in the rate of death or myocardial infarction in the CTCA group. This latter fi nding is intriguing but should be treated with caution: it was one of 22 prespecifi ed secondary endpoints, and the absolute diff erence between the two groups was only 16 events. Because randomised trials of cardiovascular imaging are fairly new, much can be learned from the well-done SCOT-HEART study to guide clinical care and future research. Many crucial questions are raised: fi rst, in the absence of a feasible placebo strategy, how can imaging studies prevent or control bias in favour of the intervention? In SCOT-HEART, CTCA created greater diagnostic certainty in CT readers than in non-reader clinicians. Second, should imaging fi ndings be prospectively linked to mandated care algorithms? The theory and practice of imaging outcomes research