Justin L. Martin, MD • Darren K. McGuire, MD, MHSc
From the Department of Internal Medicine and the Donald W. Reynolds Cardiovascular Clinical Research Center, University of Texas Southwestern Medical Center at Dallas
The Bypass Angioplasty Revascularization Investigation (BARI), funded by the National Heart, Lung, and Blood Institute (NHLBI), was designed to evaluate the long-term effectiveness and safety of percutaneous transluminal coronary angioplasty (PTCA) compared with coronary artery bypass graft (CABG) surgery for the treatment of patients with symptomatic multivessel obstructive coronary artery disease (CAD). In BARI, patients were randomly assigned to receive a primary revascularization strategy of PTCA or CABG surgery, with a primary outcome of cumulative 5-year all-cause mortality.1 The BARI results showed no difference in 5-year survival between patients who underwent CABG surgery and PTCA. However, an unanticipated finding was that in the subgroup of patients with medically treated diabetes, CABG surgery was associated with an almost 50% reduction in 5-year mortality compared with PTCA (18.4% versus 34.5%). The BARI diabetic findings prompted the NHLBI to issue an urgent “Clinical Alert” recommending CABG surgery over balloon angioplasty as the preferred revascularization strategy for the treatment of patients with diabetes and multivessel coronary disease.3 Initiated in 1991 by the National Institutes of Health to convey important research findings to the clinical arena,4 the Clinical Alert and Advisory program has successfully communi-
cated study results and recommendations through a series of 19 releases since its inception. The influence of these alerts on clinical practice remains undetermined. The goal of the present study was to use data collected from a large multicenter registry to evaluate whether the BARI diabetic findings and the subsequently released NHLBI Clinical Alert had a measurable impact on revascularization strategies (ie, PTCA versus CABG surgery) among diabetic patients with multivessel coronary disease.
Patients and methods The National Cardiovascular Network (NCN) coronary revascularization database is a collaborative project consisting of 28 hospitals from diverse geographic areas that range from modest-volume community hospitals to large-volume tertiary referral and academic medical centers. The network is thought to represent the breadth of US clinical practice. All institutions provide data on all coronary revascularizations (percutaneous and surgical) to a central data-coordinating center. We restricted our analysis to data from the 13 NCN centers that provided complete data for at least the 2 years spanning the NHLBI Clinical Alert. From these centers, we included patients with medically treated diabetes and multivessel CAD who underwent either CABG surgery or percutaneous coronary intervention (PCI) as an initial revascularization procedure and otherwise met BARI clinical inclusion criteria.
Primary data analyses were performed using multivariable logistic regression techniques that included propensity score adjustment based on likelihood to undergo PCI versus CABG surgery to account for potential differences in patient mix be-tween centers as well as over the duration of the data collection. The response variable was PCI, and the individual NCN centers were considered random effects. These methods were used to test for possible associations between time period (eg, before versus after the alert) and rate of PCI, as well as to compare PCI rates among the individual centers.
Results The study included data from 9,619 consecutive patients meeting eligibility criteria who underwent PCI or CABG surgery at one of the 13 NCN centers between 1994 and 1997. Overall, 27.5% underwent PTCA, whereas 72.5% underwent CABG surgery. Figure 1 shows the temporal trends in revascularization rates in relation to the release of the BARI Clinical Alert and the BARI trial manuscript publication. The proportion of patients undergoing PCI did not change following the Clinical Alert (28.6% before versus 26.8% after) or following the publication of the BARI manuscript (28.0% before versus 26.8% after).
The table shows the independent predictors for PCI derived from the multivariable logistic regression model. The strongest predictor of revascularization strategy used was the severity of underlying CAD. Three-vessel disease was the strongest predictor of CABG surgery, followed by two-vessel CAD that included the proximal left anterior descending coronary artery. Conversely, young age, female sex, lower Canadian Cardiovascular Society angina score, absence of heart failure, presence of renal insufficiency, less than moderate mitral regurgitation, and lack of insulin treatment were all independently associated with a higher odds ratio of receiving PCI as the initial revascularization strategy.
Although the severity of CAD was most strongly associated with revascularization strategy, no change in the rate of PCI was observed following the BARI Clinical Alert in either the subset of patients with three-vessel disease (14.9% versus 13.6%) or in the subset with two-vessel disease (49.8% versus 47.4%). When PCI use among the 13 NCN centers was analyzed, a dramatic variation in the rate of PCI was observed, ranging from 4.3% to 56.6%. This variability persisted after multivariable adjustment, even when patients were grouped according to CAD severity (25%—84% among patients with two-vessel CAD and 1%–46% among patients with three-vessel CAD). As shown in figure 2, overall PCI rates did not decrease at any of the NCN centers after the release of the Clinical Alert. In fact, PCI rates paradoxically increased at 2 of the 13 centers following the alert. As expected, the frequency of use of coronary stents increased throughout the study period, with rates being higher after, rather than before, the Clinical Alert (50.8% after versus 11.7% before; P = .001). However, no correlation between the frequency of stent use and the overall rate of PCI among the various centers was observed, suggesting that the lack of decline in overall PCI rate following the Clinical Alert was not attributable to an increased frequency of stent use at any of the individual centers.
Because the medical environment in which practitioners operate may influence clinical practice patterns, we examined the interaction between the type of institution and rate of PCI both before and after the Clinical Alert. Overall, academic medical centers were more likely than nonacademic centers to use PCI (30.0% versus 23.3%; P = .001). After adjusting for case mix, the rate of PCI did not change after the alert at either academic (30.1% before versus 32.4% after) or nonacademic centers (22.1% before versus 23.4% after). Centers formerly affiliated with the BARI trial were less likely to use PCI than nonaffiliated sites. Again, the Clinical Alert did not alter adjusted rates of PCI at either BARI sites (24.6% before versus 27.0% after) or non-BARI sites (28.7% before versus 29.5% after).
These analyses show three important findings. First, even before the BARI results were known, the majority of medically treated diabetic patients with multivessel CAD were treated with surgical coronary revascularization, especially those patients with the most advanced CAD. Second, despite the rapid and broad dissemination of the study results, the BARI diabetic findings and resultant Clinical Alert had no measurable impact on revascularization practice patterns. Finally, there is a troubling lack of clinical consensus regarding revascularization for this high-risk population, as evidenced by the greater than 10-fold variation in PCI rates among these 13 clinical centers.
The foundation of evidence-based medicine has been the randomized clinical trial, such as BARI, which, when well designed and implemented, can provide definitive comparisons of the effect of alternative treatment strategies. This allows the “evidence” to guide the management of challenging treatment dilemmas. Cardiology has been at the forefront of the evidence-based medicine movement, conducting large-scale randomized trials to address many of its most challenging treatment dilemmas. The importance of evidence-based medicine is well documented in the literature.1,5-8 Ideally, therapeutic decisions are guided by the best available evidence from clinical studies. However, the medical community has been slow to systematically incorporate the “best available data” into routine clinical practice.9,10 Some evidence suggests that the time interval between the publication of clinical trial results and the actual incorporation of new data into clinical practice may be as long as 12 years.11 How physicians respond to large-scale, multicenter, randomized trials that produce results requiring an immediate change in clinical practice has not been well studied. The real-time incorporation of clinical evidence into practice patterns remains a major challenge caused, in part, by the multitude of factors that influence clinical decision-making. Additionally, the driving forces that persuade physicians to modify their behavior are numerous. Certainly, the acceptance of various research outcomes partly depends on the quality and magnitude of the scientific evidence, timely release and dissemination of the results, relevance and generalizability of results to current practice, and incentives for adopting change in practice.5 The BARI trial diabetic results and the subsequently released NHLBI Clinical Alert provide an opportunity to examine the relationships between randomized clinical trial results and physicians’ practice patterns.
This is the first study to examine the effects of the NHLBI BARI Clinical Alert on practice patterns. On one hand, the findings are encouraging in that the vast majority of diabetic patients with multivessel coronary obstruction were treated with CABG surgery. The severity of underlying CAD was one of the most powerful predictors of undergoing CABG surgery as the initial revascularization strategy, which is consistent with evidence of benefit from prior studies.1,12,13 Despite these encouraging findings, it remains disappointing that neither the NHLBI Clinical Alert nor the subsequent publication of the article on the BARI diabetic findings changed clinical practice. Although we observed no change during our study period in the proportion of PCIs, other data have shown a 20% increase in these procedures as a percentage of all revascularizations in the United States over the past decade.14 Therefore, the absence of an increase in PCI rates may truly represent the clinical influence of the BARI diabetic findings.
Perhaps the most powerful finding of our study was the clear lack of clinical consensus regarding the best initial revascularization strategy for diabetic patients with multivessel CAD. This lack of consensus was highlighted by the over 10-fold variation in PCI use among the 13 US centers we included in our study. This lack of consensus does not appear to derive from a lack of awareness of the BARI diabetic findings, however, based on our survey of the principal investigators at each of the participating NCN sites. Our survey showed that the BARI diabetic findings were broadly disseminated and were considered to be valid and generalizable to patients undergoing treatment at the participating hospitals. Despite these re-sults, the respondents uniformly conceded that practice patterns at their respective institutions had not materially changed in the wake of the BARI diabetic findings and NHLBI Clinical Alert, with most attributing the disconnect to advances in technologies and adjunctive therapies that seemed to render the BARI findings obsolete. However, even this notion was countered by our analyses, which showed no correlation between frequency of intracoronary stent use and the proportion of PCIs performed at individual centers. Therefore, the explanation for the widely varied clinical practice remains unknown.
ConclusionOur findings underscore the gap that often exists between well-conducted scientific clinical evidence and the day-to-day care of individual patients. The inability of the Clinical Alert and the BARI trial diabetic findings to measurably influence clinical practice patterns highlights the multitude of challenges involved in implementing the ra-
pid accumulation of evidence-based medicine into the daily practice of the medical community. Clearly, the ability of large-scale multimillion-dollar clinical trials, such as BARI, to influence therapeutic decision-making is complex. The quality of available evidence and its timely dissemination to practitioners represents only a portion of this puzzle. Further study of how we can best incorporate clinical trial results into medical practice is warranted so that we continue to consistently apply the best available evidence to the treatment of the cardiovascular patients receiving our care.