Three-year Survival after Percutaneous Coronary Intervention According to Appropriateness Criteria for Revascularization
- Volume 21 - Issue 11 - November, 2009
- Posted on: 11/5/09
- 0 Comments
- 8341 reads
ABSTRACT: Objectives. We sought to compare 3-year outcomes of percutaneous coronary intervention (PCI) according to recently published appropriateness criteria for PCI. Background. The choice of revascularization between PCI and coronary artery bypass grafting (CABG) remains uncertain in many patients despite numerous randomized clinical trials and meta-analyses. Methods. Consecutive patients undergoing a first PCI at a single, large-volume institution were included if they did not have prior CABG and did not need emergency PCI. Patients were classified according to PCI indication into the following groups: Appropriate (A) – 1- or 2-vessel coronary artery disease (CAD), Uncertain (U) – 3-vessel CAD and Inappropriate (I) – left main coronary artery stenosis. Survival was assessed with the Social Security Death Index. Results. A total of 2,134 patients fulfilled the study criteria: 1,706 (80%) with “appropriate” PCI, 414 (19.4%) with “uncertain” PCI and only 14 (0.6%) with “inappropriate” PCI. In-hospital outcomes were very favorable, with 99.3%, 98.6% and 100% of the three groups, respectively, experiencing no complications (p = 0.31). The estimated survival in the three categories at 900 days was 92.6% (95% confidence interval 91–94%) for Group A, 91.3% (88–4%) for Group U and 66.9% (33–87%) for Group I; p = 0.014. The only predictors of mortality were advanced age and comorbidities, but not “appropriateness level” (p = 0.26). Conclusion. The majority of PCIs performed would were classified as “appropriate.” The patients classified as “uncertain” had similarly favorable outcomes, as those considered “appropriate” both during initial hospitalization and during the 3-year follow up. If confirmed, these data suggest that anatomically-based appropriateness criteria are not sufficient to inform choice of revascularization method.
J INVASIVE CARDIOL 2009;21:554–557
Key words: revascularization, PCI, appropriateness, survival
Coronary revascularization appropriateness has been recently reviewed by a panel of 17 experts in adult cardiovascular disorders, internal medicine and cardiothoracic surgery.1 Each member assigned an appropriateness score (1–9) for approximately 180 common clinical scenarios, which considered clinical presentation, severity of angina, extent of ischemia on noninvasive testing, intensity of medical therapy and anatomical extent of coronary disease (CAD). In a subset of these scenarios, assuming that revascularization was appropriate, consideration was given to the appropriateness of percutaneous coronary intervention (PCI) or coronary artery bypass grafting (CABG) for that scenario, where each method was considered separately, and not one against the other. The level of appropriateness was graded as appropriate (score 7–9), uncertain (4–6) and inappropriate (1–3), based on expert consensus, using a modified Delphi exercise described by RAND.1
We sought to analyze, according to these new appropriateness criteria for PCI, the outcomes of patients selected for revascularization and undergoing PCI at a single institution over a 3-year period.
Using the mandatory New York State PCI registry reporting forms (versions of DOH 3331 from November 2004, January 2006 and January 2007), we classified all PCI procedures performed in 2005–2007 according to appropriateness versus CABG based on the anatomical extent of CAD.1 In brief, PCI was considered inappropriate for all procedures involving the left main coronary artery (LMCA), uncertain in patients with 3-vessel CAD and appropriate in all the others. A vessel was considered diseased if there was at least one stenosis ≥ 70% (50% for the LMCA). Patients with a LMCA stenosis were classified as having 2- or 3-vessel CAD depending on coronary dominance and involvement of the right coronary artery. Patients were included if they did not undergo previous CABG, did not need emergency PCI, did not have ST-elevation myocardial infarction (MI) within the past 12 hours and were not hemodynamically unstable. For each patient having more than one PCI during the study period, only the first was considered for outcome analysis. The primary endpoint of the study was all-cause mortality ascertained using the Social Security Death Index (SSDI).2 Additional endpoints were in-hospital death, emergency CABG or Q-wave infarct. We excluded patients who did not fit the above criteria and those without a valid social security number.
All variables common to the different versions of the PCI report form were used to build the dataset. Statistical analysis included analysis of variance and the chi-square test for comparison of continuous and categorical variables, as appropriate. Determination of vital status was made by querying the SSDI (updated on February 17, 2009) between April 6, 2009 and April 8, 2009. Survival analysis was performed using Kaplan-Meier methodology with testing. A multivariable Cox proportional hazard model was developed to assess the impact on all-cause mortality of the following variables: appropriateness of PCI, age, gender, race, PCI on day of angiography, prior PCI, priority of PCI (elective vs. urgent), height and weight, ejection fraction, creatinine level, angina class, previous infarct, cerebral or peripheral arterial disease, heart failure (present or past), chronic lung disease, diabetes, end-stage renal failure, presence of stent thrombosis, number of diseased vessels and number of lesions treated. P-values < 0.05 were considered statistically significant. All analyses were performed using STATA SE 9.0 (StataCorp, College Station, Texas).
There were 3,375 PCI procedures performed between January 1, 2005 and December 31, 2007. Of these, 95 patients had missing or incorrect social security numbers, 444 patients had previous CABG, 114 patients underwent emergency PCI or were hemodynamically unstable and 588 cases (in 483 patients) were repeat procedures. After these exclusions, 2,134 patients undergoing a first PCI procedure with stable CAD and without previous CABG were included (Table 1). The ejection fraction was ≤ 30% in 10% of patients, diabetes mellitus was present in 39% and nearly two-thirds were > 65 years of age.
There were 1,706 patients with “appropriate” PCI (A: 80%), 414 with “uncertain” PCI (U: 19.4%) and only 14 with “inappropriate” PCI (I: 0.6%). In-hospital outcomes were similar among the 3 groups (Table 2). There were no strokes, emergency CABG or Q-wave MI prior to hospital discharge. The number of lesions treated in patients with 1, 2 and 3-vessel CAD was 1.3 ± 0.7, 1.9 ± 1.1 and 2.2 ± 1.4, respectively.
The mean follow-up period was 915 ± 341 days and all survivors were followed for at least 14 months. One hundred forty-six deaths occurred: 111, 31 and 4 in groups A, U and I, respectively. The estimated survival in the three categories at 900 days was 92.6% (95% confidence interval 91–94%) for Group A, 91.3% (88–94%) for Group U and 66.9% (33–87%) for Group I; p = 0.014 (Figure 1). The independent predictors of survival are shown in Table 3. The level of appropriateness did not correlate with survival (p = 0.26). The principal determinants of increased mortality were increased age, congestive heart failure and comorbidities, such as cerebral or peripheral disease and dialysis. Non-Caucasians had better outcomes. The proportional hazards assumption was met (p = 0.55, Schoenfeld residual method). The model had a satisfactory discriminatory ability (C-statistic = 0.76). We repeated this analysis after exclusion of the 14 patients in Group I. There was no material change in results. The level of PCI “appropriateness” remained unrelated to outcome, p = 0.86.
There were 320 patients among group A (18.8%) with 2 or more procedures (total of 371). Fifty of the 414 patients in Group U had ≥ 2 procedures, 50% of which were staged for multivessel CAD. In 113 patients, the indication changed among the repeat procedures, between Groups A and U, depending on the number of diseased vessels and patency of previous PCI sites. Only in two Group U patients did the indication change to “inappropriate.” Among Group I patients, there were no repeat PCI procedures. The EUROSCORE3 among patients with “inappropriate” PCI, who would have been candidates for CABG based on their anatomy, was 4.8 ± 2.9. The predicted in-hospital mortality rate for such a cohort of patients is 3.0% (2.6–3.5%); it was 0% in these PCI patients. Two of these patients were 90 and 91 years of age, and two others were in their ninth decade.
The principal findings of our study are that in a combined academic and private practice of interventional cardiology, 80% of PCIs performed would be classified as “appropriate” according to recently published criteria. The other 20% of patients with “uncertain” indications were treated with a percutaneous approach for 3-vessel CAD, and their outcome, specifically all-cause mortality, was very similar to that of the “appropriate” group. Furthermore, in multivariable analysis, the criteria for appropriateness are not related to outcomes, which is principally dictated by age and serious comorbidities. The number of patients in Group I was too small for meaningful analysis and conclusions. To our knowledge, this is the first study to assess the relation between criteria for PCI appropriateness and long-term outcomes.
Besides the coronary anatomy, many other considerations are involved such as advanced age, renal dysfunction, number and characteristics of lesions4 and patient preference. In the recently published Synergy between Percutaneous Coronary Intervention with Taxus and Cardiac Surgery (SYNTAX) trial,5 patients with 3-vessel CAD or LMCA stenosis (groups U and I in our study) randomized to PCI or CABG had similar rates of death at 1 year (4.3% vs. 3.5%, respectively; p = 0.37), but markedly different rates of repeat procedures (favoring CABG) and stroke (favoring PCI). In our study, patients in Group U had an estimated mortality of 4.5% at 1 year, despite being 5 years older and having diabetes nearly twice as often compared with patients in SYNTAX.