The Appropriate Use Criteria for Revascularization (AUC) is a vital instrument supporting high-quality interventional practice that presents, in a pragmatic format, the optimal application of the evidence base and existing guidelines to patient selection for percutaneous coronary intervention (PCI). It is a meaningful response to concerns of overutilization, serves as a quality metric, and establishes the relative roles of coronary artery bypass graft surgery (CABG) and medical therapy in patients with both stable ischemic heart disease (SIHD) and acute coronary syndromes (ACS). The original AUC¹ and a focused update² set the standard of practice regarding optimal patient selection for PCI. The AUC has broad implications for the future of cardiovascular health care, especially in regard to case selection, treatment options, and reimbursement. In the future, it will become the foundation for tethering payment decisions and quality assessment to patient-centered therapeutic decision making. Limitations of the first version have been extensively articulated, and a subsequent revision has been published improve the criteria so they more meticulously take into account how interventional cardiologists evaluate scientific data and apply those conclusions to individual revascularization decisions.

Today’s generation is the recipient of a myriad of astonishing bioengineering, pharmacologic, and technical innovations over the last 3 decades. These advancements have resulted in the growth and maturity of a new field of treatment for coronary artery disease (CAD). The exponential growth in the number of PCI procedures has resulted in an unprecedented escalation in the cost of health care: approximately 600,000 PCIs are performed in the United States each year³ at a cost exceeding $12 billion. However, this expense is not accompanied by a measurable outcomes benefit.⁴ Moreover, the wide geographic variability of the use of PCI in the United States^5,6 has been interpreted as a demonstration that patients are not being referred for the procedure strictly based on scientific evidence. The divergence in cost versus benefit logically demands a formal evaluation of PCI utilization.

The value of PCI in ACS (ST-segment elevation myocardial infarction [STEMI], non–ST-segment elevation myocardial infarction [NSTEMI], and unstable angina) has been definitively demonstrated: by eliminating the causative coronary obstruction, PCI significantly reduces mortality and recurrent myocardial infarction in this setting. However, in SIHD, there is no proven survival benefit or reduction in subsequent myocardial infarction versus medical therapy. PCI in this circumstance reduces or relieves angina but does not contribute to improvement in hard end points. Furthermore, since patients who undergo PCI are exposed to the risk of periprocedural complications, including death, stroke, bleeding, and myocardial infarction, a judicious approach to case selection is required. Deciding which patients are likely to benefit from PCI is as much a matter of clinical judgment as the application of clinical evidence (clinical trials, registries, meta-analyses, retrospective studies confirmed by additional data).⁷ In contemporary practice, the decision of whether PCI, medical therapy, or bypass surgery should be the primary strategy in a given patient can no longer be made based simply on the number of vessels diseased, as it was in the past.⁸ Not every patient with a hemodynamically significant stenosis, a positive stress test, or angina experiences improved survival or better angina relief from PCI compared to other treatment modalities. For example, in patients with SIHD, PCI provides only a modest population-average improvement in symptom relief compared with medical therapy.⁹ No patient wants to endure the pain and recuperative process of surgery if there is no incremental benefit; conversely, no patient will accept the cost and hassle of multiple medications and their adverse effects if there is a better alternative. Further, it is reasonable to avoid expensive procedures unless there is some measurable improvement over other options. Because it is not possible to quantitatively scale the expected advantages versus the inherent drawbacks and risks, which ultimately is an individual subjective determination, there is no objective way to compare the different rates of success and adverse events of each treatment method; thus, clinical judgment requires an individual assessment in every case.

It is important that expert practitioners set parameters and then determine the extent to which PCI is performed for appropriate versus inappropriate indications to identify procedural overuse and areas for quality improvement and cost savings. Therefore, a means of evaluating the clinical appropriateness of PCI is necessary; good outcomes and high operator volumes alone are not sufficient to define high quality. These surrogate measures in isolation encouraged the performance of large volumes of low-risk procedures of uncertain appropriateness and the avoidance of high-risk but properly selected cases.¹⁰ Concern regarding overutilization is a crucial issue the profession must address if it is to continue to enjoy the privileges of self-referral and self-regulation. Other potential advantages of specific criteria to guide appropriate case selection are also apparent, including providing guidance for clinicians and patients in making difficult decisions, enabling comparison of a laboratory’s or operator’s outcomes versus a national benchmark, assisting in providing the “right therapy at the right time,” and upholding a patient-centered approach.

In contrast to practice guidelines,¹¹ which summarize published data to classify levels of utility, or performance measures,¹² which distill the guidelines into specific actions that should be taken programmatically, the AUC integrate the guidelines, clinical trial evidence, and clinical experience to specify what clinical decision is usually optimal.¹³ The AUC define the strength of the indications for a procedure, weighing both its benefits and drawbacks. The goal of the AUC is to guide decision making by clarifying expected benefits in specific clinical situations. In 2009, the AUC for coronary revascularization were published,¹ and in 2011, rates of appropriateness in the National Cardiovascular Disease Registry (NCDR) were reported.¹⁴ A focused update reflecting updated comparative data between bypass surgery and PCI was published in 2012.² Another update was published after this chapter was submitted. It contains substantial changes that will be reflected in the online version of this book.

Ethical decision making requires of every physician that procedures should not be performed unless they offer potential benefit to the patient; this principle has been an accepted tenet of medicine since the Hippocratic Oath. How to define benefit and how clinical decisions should be evaluated in this context are constantly changing features of modern medicine. In regard to PCI, the initial notion underlying balloon angioplasty was that the treatment of any single high-grade coronary stenosis that was technically feasible offered benefit over medications. In 2006, the OAT trial demonstrated that not all recent occlusions opened percutaneously had improved outcomes, and this provocative trial was followed the next year by the COURAGE trial,⁹ which suggested that optimal medical therapy in multivessel SIHD may produce similar hard end point outcomes (death, myocardial infarction, and stroke) as complete revascularization with stenting. With the development of drug-eluting coronary stents and improved catheter and guide wire technology, it has become even more ambiguous which patients are best treated with pharmacology, bypass surgery, or a percutaneous procedure. Consequently, the nuances and complexity of determining when a PCI is and is not appropriate is a controversial and active area of investigation.

The first large-scale investigation of an evidence-based classification to define appropriateness studied the relationship between the American College of Cardiology (ACC)/American Heart Association (AHA) recommended indications for PCI and short-term in-hospital outcomes and was performed using data from the NCDR.¹⁵ After excluding STEMI, the remaining 412,617 PCIs were grouped by indication class: class I, evidence and/or agreement that PCI is useful and effective; class IIa, conflicting evidence and/or divergent opinions, weight is in favor; class IIb, usefulness/efficacy is less well established; and class III, evidence and/or agreement that PCI is not useful or effective and may be harmful. Frequency of indications was as follows: class I, 64%; class IIa, 21%; class IIb, 7%; and class III, 8%. Clinical success declined across the indication classes (92.8%, 91.7%, 89%, and 85.5% for class I, IIa, IIb, and III, respectively; P < .001), whereas adverse events increased. The results of this study demonstrated that most procedures were performed for class I indications and that a significant relationship existed between evidence-based indications recommended by the ACC/AHA Task Force and in-hospital outcomes. The authors concluded that closer adherence to guidelines can reduce variations in care and improve quality and may ultimately result in better outcomes.

In a follow-up study,¹⁶ the outcomes in each indication class were evaluated after risk adjustment by the NCDR risk adjustment model. A total of 559,273 PCI procedures were analyzed. Increasing frequencies of risk components were observed across classes I, IIa, IIb, and III. Expected mortalities for each class calculated by the risk adjustment model were close to observed values (expected 0.52%, 0.59%, 1.72%, and 1.96%, respectively; observed 0.49%, 0.63%, 1.88%, and 1.60%, respectively). Thus, the ACC-NCDR risk-adjusted mortality model, when linked to the ACC/AHA PCI guidelines, produced mortality risk estimates by indication class close to actual observed values. The critical implication of this finding was that the model, together with the PCI guidelines, could be used as a powerful analytic tool for quality assurance and assessment of the appropriateness of case selection. Indeed, the recognition that the flipside of the development of an accurate risk prediction model for PCI is its application to quality assurance was the crucial step in the evolution to AUC.

The essential rationale for the AUC is to provide an evidence-based, clinically applicable assessment of procedural appropriateness without regard to financial considerations or socioeconomic status. The ACC convened a committee to develop AUC derived from the PCI guidelines. The process chosen combined evidence-based medicine, guidelines, and practice experience by engaging a technical panel in a modified Delphi exercise as previously described by RAND.^1,17 The technical panel was composed of 4 interventional cardiologists, 4 cardiovascular surgeons, 8 general cardiologists and other physicians who treat patients with cardiovascular disease or health outcome researchers, and 1 health plan medical officer (also a physician). One hundred eighty scenarios illustrative of commonly observed clinical situations were developed by a writing committee and scored by the separate technical panel on a scale of 1 to 9. Scores of 7 to 9 indicated that revascularization was considered “appropriate” and likely to improve health outcomes or survival. Scores of 1 to 3 indicated revascularization was considered “inappropriate” and unlikely to improve health outcomes or survival. The mid-range (score of 4-6) indicated a clinical scenario for which the likelihood that coronary revascularization would improve health outcomes or survival was considered “uncertain.” These categories are those mandated by the RAND process. For the majority of the clinical scenarios, the panel only considered the appropriateness of revascularization irrespective of whether this was accomplished by PCI or CABG. In a select subgroup of clinical scenarios in which revascularization was generally considered reasonable, the appropriateness of PCI and CABG individually as the primary mode of revascularization was considered. Because SYNTAX¹⁸ had not been published in a peer review format at the time of the meeting of the technical panel (November 2008), the initial document¹ was written explicitly to concur with the existing guidelines (Figs. 66-1 and 66-2). A revised focused update was later published² that incorporated the SYNTAX results (Fig. 66-3).

The decision to perform coronary revascularization is correctly based on a multitude of factors, including the presence of a stenosis-producing ischemia, the severity of symptoms, application of medical therapy, objective evidence of ischemia, the volume of myocardium at jeopardy, the presence of viable myocardium with the potential for reversible dysfunction, and the demonstration of significant coronary stenoses that are treatable with current techniques. Each of these conditions requires substantial testing to demonstrate, and their interrelationships are notably variable.

The failure to find a quantitative method that resolves the uncertainty that attends the subjective interpretation of angiograms impacts the selection of patients for interventional procedures. Coronary angiography is widely recognized to have many imprecisions, especially with a stenosis in the intermediate range, yet it remains the primary diagnostic tool. Adding a physiologic measure of stenosis severity to angiographic assessment enhances the accuracy of diagnosis and optimizes the selection of patients for expensive and risky treatment options. The structure of the AUC is predicated on an assumption that the size of a defect on nuclear scans can be used as the standard for assessing the significance of a stenosis and the amount of myocardium subtended by a critical stenosis.

For purposes of describing the severity of symptoms, the Canadian Classification was used. To define optimization of medical therapy, an assumption was made that at least 2 antianginal drugs constituted a pragmatic clinical definition of “optimal medical therapy,” as described in COURAGE.¹⁹

Two different formats were used to summarize the voting results. An algorithmic or decision-tree approach was chosen for the ACS indications, while a matrix method was chosen to summarize the SIHD portion.

A multicenter, prospective study of all PCI patients within NCDR undergoing PCI between July 1, 2009, and September 30, 2010, at 1091 US hospitals was undertaken.¹⁴ The appropriateness of PCI was adjudicated using the AUC. Of 500,154 total PCIs, 355,417 (71.1%) were for ACS (STEMI, 103,245 [20.6%]; NSTEMI, 105,708 [21.1%]; high-risk unstable angina, 146,464 [29.3%]), and 144,737 (28.9%) were for SIHD. In ACS, 350,469 PCIs (98.6%) were classified as appropriate, 1055 (0.3%) as uncertain, and 3893 (1.1%) as inappropriate. For SIHD, 72,911 PCIs (50.4%) were classified as appropriate, 54,988 (38.0%) as uncertain, and 16,838 (11.6%) as inappropriate. The majority of inappropriate PCIs for SIHD were performed in patients with no angina (53.8%), low-risk ischemia on noninvasive stress testing (71.6%), or suboptimal (≤1 medication) antianginal therapy (95.8%). Furthermore, although variation in the proportion of inappropriate PCI across hospitals was minimal for acute procedures, there was substantial hospital variation for PCI in SIHD (median hospital rate for inappropriate PCI, 10.8%; interquartile range, 6.0%-16.7%). Overall, the rate of inappropriate PCI was 4.1%; hospitals in the lowest quartile had inappropriate PCI rates of ≤1.73% versus ≥4.62% for the highest quartile hospitals.

The New York State Registry corroborated these findings generally in patients with SIHD.²⁰ In this mandatory state registry, 14% of PCI patients were found to have inappropriate AUC classification, confirming that a small but definite incidence of inappropriate PCI is not a spurious observation. Furthermore, 28% of all PCIs performed in the state lacked sufficient noninvasive test information to be rated. Most concerning was the relatively low percentage (36.1%) of patients undergoing PCI classed as appropriate by AUC.

Bradley et al²¹ evaluated the association between patient selection for PCI and postprocedural outcomes. In 203,531 patients undergoing elective PCI, the association between a hospital’s proportion of nonacute PCIs categorized as inappropriate by the 2009 AUC and in-hospital mortality, bleeding complications, and use of optimal guideline-directed medical therapy at discharge (ie, aspirin, thienopyridines, and statins) was evaluated. When categorized as hospital tertiles, the range of inappropriate PCI was 0.0% to 8.1% in the lowest tertile, 8.1% to 15.2% in the middle tertile, and 15.2% to 58.6% in the highest tertile. Compared with lowest-tertile hospitals, mortality was not significantly different at middle-tertile (adjusted odds ratio [OR], 0.93; 95% confidence interval [CI], 0.73-1.19) or highest-tertile hospitals (OR, 1.12; 95% CI, 0.88-1.43; P = .35 for differences between tertiles). Similarly, risk-adjusted bleeding did not vary significantly (middle-tertile OR, 1.13; 95% CI, 1.02-1.16; highest-tertile OR, 1.02; 95% CI, 0.91-1.16; P = .07 for differences between tertiles) nor did use of optimal medical therapy at discharge (85.3% vs 85.7% vs 85.2% for lowest, middle, and highest tertile, respectively; P = .58). Thus, a hospital’s proportion of inappropriate PCIs was not associated with in-hospital mortality, bleeding, or medical therapy at discharge. This finding suggests PCI appropriateness measures aspects of hospital PCI quality that are independent of operator or technical quality. Therefore, PCI appropriateness and postprocedural outcomes are both important metrics of PCI quality. However, PCI appropriateness was not associated with in-hospital survival.

In a follow-up study,²² the demographics underlying inappropriate PCI as defined by the AUC were evaluated. Of 211,254 nonacute PCIs, 25,749 (12.2%) were classified as inappropriate. After multivariable adjustment, men (adjusted OR, 1.08; 95% CI, 1.05-1.11; P < .001) and whites (adjusted OR, 1.09; 95% CI, 1.05-1.14; P < .001) were more likely to undergo an inappropriate PCI in comparison with women and nonwhites. Compared with privately insured patients, those who had Medicare (adjusted OR, 0.85; 95% CI, 0.83-0.88), other public insurance (adjusted OR, 0.78; 95% CI, 0.73-0.83), and no insurance (adjusted OR, 0.56; 95% CI, 0.50-0.61) were less likely to undergo an inappropriate PCI (P < .001). In addition, compared with urban hospitals, those admitted at rural hospitals were less likely to undergo inappropriate PCI, whereas those at suburban hospitals were more likely. Thus, for nonacute indications, PCIs categorized as inappropriate were more commonly performed in men, whites, and patients with private insurance. The authors concluded that higher rates of PCI in these patient populations might either reflect potential overuse in privileged patients or underuse in disadvantaged populations. Both are problematic: inappropriate procedures expose the patient to unnecessary risk, while underuse in disadvantaged patient groups leads to disparities in care.

Unfortunately, the AUC and its use as a quality metric have become sources of controversy and dispute. Pervasive media attention was unleashed with publication of the NCDR analysis that labeled 4.1% of PCIs in the United States as “inappropriate.” The media, and much of the lay public, were especially drawn to the finding that 11.6% of nonacute, elective PCIs were classified as inappropriate and, additionally, to the wide variation in the rates of inappropriate nonacute PCI across hospitals (0%-50%). These circumstances led to claims that billions of dollars are being spent on completely unnecessary procedures. Superimposed on widely reported fraudulent behavior on the part of a very few but high-profile PCI operators, widespread concern was raised.