LAUSR

Care providers faced with a patient with chest pain have a number of options for noninvasive testing. They also have the option of ‘‘not testing’’ in patients with a low clinical likelihood of angina and a low-risk profile. A strong scientific basis to justify the use of one test over others is lacking and difficult to procure given the magnitude of the undertaking required. Diagnostic tests are usually compared using metrics of test efficacy (diagnostic accuracy). However, metrics of test effectiveness may be equally relevant, including patient outcome after a test, test effectiveness in the real world, downstream resource utilization, and societal cost. In this issue of the journal, Qinli Ma and colleagues report on their analysis of a large dataset of Medicare patients, which was designed to compare patient outcomes (unstable coronary syndromes, all-cause and cardiac hospitalizations, and all-cause death), downstream resource utilization (CT coronary angiography, invasive angiography, PCI, or CABG), and cost (cardiac and all-cause medical costs reported per patient per month) between patients with suspected coronary artery disease (CAD) who underwent clinically indicated positron emission tomography (PET) myocardial perfusion imaging (MPI) without absolute flow quantification or conventional testing (stress echocardiography or SPECT MPI; notably, patients who had exercise stress without imaging or CT coronary angiography were not included) as the first step in their evaluation. The question is relevant because PET has established advantages over SPECT including a superior diagnostic accuracy for obstructive coronary artery disease (CAD), but is also more expensive. Furthermore, based on its advantages over SPECT, societal position papers have recommended PET as the preferred test for patients with suspected CAD who are unable to exercise (taking into consideration the generally accepted notion that functional parameters derived from exercise testing provide incrementally useful information and should be obtained when possible, thus making SPECT the preferred modality in such patients). The authors analyzed data from a nationally representative random sample of 5% of Medicare Fee for service subjects registered with a single large US health insurance provider (Anthem, Inc), who had undergone index imaging for chest pain during a defined 4-year period. Using a standard statistical approach to minimize differences between groups in retrospective data analysis, the authors applied a propensity score to match 4691 patients who underwent PET (3.2% of sampled patients) to an equal number of patients who underwent conventional testing (matched from a total of 139, 884 patients). A propensity score is the probability that a subject with certain characteristics will be assigned to one of the two groups. The investigators used a multivariable logistic regression analysis to estimate the propensity score (likelihood) of having PET as the first imaging test (discriminant analysis and random forests are other potential approaches for deriving a propensity score) and used a ‘‘greedy’’ algorithm to create a one-to-one matched group (as opposed to matching each subject from one group to many subjects in the other group) sampled without replacement (each subject used in only one pairing). In the greedy matching algorithm used, the subject with the Reprint requests: Prem Soman, MD, PhD, FRCP (UK), FACC, MASNC, Division of Cardiology and The Heart and Vascular Institute, University of Pittsburgh Medical Center, A-429 Scaife Hall, 200 Lothrop Street, Pittsburgh, PA 15213; [email protected] J Nucl Cardiol 2021;28:2138–40. 1071-3581/$34.00 Copyright 2020 American Society of Nuclear Cardiology.