Coronary artery disease is the single largest killer of American men and women (1). In 2004 in the U.S., there were 840,000 cases discharged with the diagnosis of acute coronary syndrome, most of them with acute myocardial infarction (1). In the last decades, important advances have been made, which have provided patients with coronary artery disease with effective drugs able to improve significantly their prognosis, such as antiplatelet agents, statins, beta-blockers, and angiotensin-converting enzyme inhibitors (2- 4).

Percutaneous coronary interventions (PCIs) are increasingly being used in patients with various manifestations of coronary artery disease. They represent an established treatment strategy that improves survival and survival free of recurrent myocardial infarction in patients with ST-segment elevation myocardial infarction (5- 6). Early invasive therapy also improves long-term survival and reduces late myocardial infarction in patients with non–ST-segment elevation acute coronary syndromes (7). Although PCI reduces symptoms in patients without acute coronary syndromes (8), its effects on the prognosis of these patients are still not defined. The assessment of this issue has been difficult for at least 2 reasons. First, patients with stable coronary artery disease have a very good prognosis and large sample size studies are required to assess potential differences in treatments regarding rare events (9- 10). All studies performed to date were far from having sufficient power to assess mortality. Second, there is a certain risk associated with PCI, which leads to aggregation of events in a relatively short period after the procedure. Any potential beneficial effect of PCI compared with medical treatment alone may require time to offset this early excess risk, so that an extended follow-up may enable a more unbiased evaluation of the relative merits of these treatment strategies.

In an attempt to overcome these difficulties, 2 meta-analyses have been performed by pooling together the results from several randomized trials comparing invasive and medical treatment strategies in patients with no acute coronary syndromes (11- 12). The first meta-analysis summarized the results of 6 trials including 1,904 patients, and the analysis of mortality was based on 26 death cases (11); the second meta-analysis summarized the results of 11 trials including 2,950 patients, and the analysis of mortality relied on 196 death cases (12). Even in a recent brief update, several randomized clinical trials with a design similar to those already included were left unaccounted for in the latter meta-analysis (13). This underscores the need for a new, comprehensive meta-analysis of all pertinent studies on the impact of PCI on the prognosis of patients with stable coronary artery disease.

The purpose of this meta-analysis was to evaluate whether PCI affects the long-term prognosis of patients with stable coronary artery disease.

We intended to retrieve all randomized trials comparing a PCI-based invasive treatment strategy with a medical treatment strategy in patients with coronary artery disease and symptoms or signs of ischemia. Trials that included patients with acute coronary syndromes (with or without ST-segment elevation on electrocardiogram, with or without troponin or cardiac enzyme elevations) within the first week from presentation were excluded from this meta-analysis. The search was performed for the period between January 1, 1980, through August 20, 2007, and involved the PubMed database, U.S. National Institute of Health, Cochrane Central Register of Controlled Trials, proceedings of the American Heart Association, American College of Cardiology, and European Society of Cardiology, as well as internet-based sources of information on results of clinical trials in cardiology. Other data sources included reference lists of retrieved articles, and pertinent reviews and editorials from leading medical journals.

Nineteen randomized clinical trials were identified that assigned patients with no acute coronary syndromes to an invasive, PCI-based treatment strategy or medical-only treatment strategy (14- 32). In 2 of these trials (22,31), neither symptoms nor signs of myocardial ischemia were a prerequisite for enrollment of patients in the study; thus, they were excluded from the present meta-analysis. Of the 17 trials included, 1 was published in abstract form (29) and 16 as full articles mostly presenting updated, extended follow-up (14- 21,23- 28,30,32). In 2 trials (23,30), patients were randomly assigned to 1 of 3 treatment groups: PCI, medical treatment, or coronary artery bypass grafting (CABG); for both of these trials, the CABG treatment arm was not included in this meta-analysis. However, patients who were randomly assigned to the PCI-based strategy or medical treatment group but received CABG were not excluded from this meta-analysis.

The primary end point of this meta-analysis was all-cause death within the longest follow-up period that was published by the investigators. Other outcomes of interest were death attributable to cardiac causes and myocardial infarction. The definition of the end point of myocardial infarction is shown in (Table 1).

All trials included in this meta-analysis were prospective, randomized trials. The existence of an independent event committee was declared by the investigators of 12 of 17 trials (14- 19,21,24- 26,28,32). Baseline characteristics were evenly distributed between the 2 treatment groups in all trials. The actual treatment received was clearly shown for all trials. In all but 2 trials (16,27), the analysis was performed on the basis of the intention-to-treat principle. In the 2 trials that did not follow this principle, a total of 5 patients were excluded after randomization (16,27).

Treatment effect could not be assessed for trials in which the event of interest was not observed in any of the treatment groups. For trials in which only 1 of the treatment groups had no events of interest, the treatment effect estimate and its standard error were approximated from 2 × 2 contingency tables after adding 0.5 to each cell (33). We used the Cochran Q-test to assess heterogeneity across trials. Also, we calculated the I² statistic to measure the consistency between trials with values of 25%, 50%, and 75% defining the cutoff points for identifying low, moderate, and high degrees of heterogeneity, respectively (34). Treatment effects from individual trials were pooled using both the fixed effects Mantel-Haenszel model (35) and the random effects DerSimonian and Laird model (36). Several additional analyses were carried out to assess potential bias regarding the primary end point of the study. First, sensitivity analyses were performed by comparing the treatment effects obtained with each trial removed consecutively from the analysis with the overall treatment effects. Second, a funnel plot as well as the Egger test were used to assess publication bias (37). Third, a funnel plot was constructed to graphically illustrate the relationship between treatment effect size and sample size of the individual trials. Finally, we used a random-effects meta-regression analysis to estimate the extent to which other covariates—year of completion of patient enrollment, total sample size of the trial, proportion of patients with previous myocardial infarction, proportion of patients who received stents or CABG in the PCI group, proportion of patients who received non-protocol revascularization in the medical treatment group—might have influenced the treatment effect. All p values are 2-sided. Statistical significance was assumed for p < 0.05. Statistical analysis was performed using the Stata software, version 9.2 (Stata Corp., College Station, Texas).

(Table 1) shows inclusion and exclusion criteria as well as the primary end point in individual trials. A total of 17 randomized trials including 7,513 patients were analyzed. (Table 2) shows the main characteristics of the patients in each trial. Overall, the average age of the patients was 60 years, 18% of them were women, 54% had incurred myocardial infarction, and the average length of follow-up was 51 months. Ninety-two percent of the patients in the PCI-based strategy group received revascularization (43% balloon angioplasty, 41% stents, and 8% CABG). Drug-eluting stents were only used in 31 patients in 1 trial (15). In the medical treatment group, 28% of the patients received nonprotocol revascularization early or at some point in time during follow-up.

In the PCI group, 271 patients died, compared with 335 patients in the medical treatment group. Allocation to the PCI group was associated with a 20% reduction in the odds ratio (OR) of all-cause death (Figure 1). There was no inconsistency across the trials (I² = 17%). The sensitivity analysis yielded ORs that ranged from 0.76 (95% confidence interval [CI]: 0.60 to 0.96) to 0.86 (95% CI: 0.72 to 1.03) that were not significantly different from the overall OR (p ≥ 0.591). (Figure 2)A shows the funnel plot of publication bias, which was not statistically significant (p = 0.261) on the basis of the Egger test. (Figure 2)B shows that there was no relationship between sample size and treatment effect size. None of the covariates—year of completion of patient enrollment (p = 0.982), total sample size of the trial (p = 0.634), proportion of patients with previous myocardial infarction (p = 0.119), proportion of patients in the PCI group who received stents (p = 0.9361) or CABG (p = 0.392), and proportion of patients in the medical treatment group who received nonprotocol revascularization (p = 0.652)—showed a significant interaction with treatment effect.

A separate analysis was performed to investigate the role of length of follow-up. We assessed all-cause death after follow-up periods of up to 1 year, 3 years, and 5 years, and added these data to our overall analysis, encompassing a follow-up period of up to 10 years. For this purpose, we sought additional information on earlier mortality in the trials from initial publications (38- 42) or directly from investigators (18). (Figure 3) shows the relationship between the length of follow-up and the ORs of all-cause death associated with PCI versus medical treatment.

Patients with acute coronary syndromes were excluded from the studies included in this meta-analysis. However, all but 1 trial (23) had included patients with previous myocardial infarction in a proportion ranging from 25% to 100%. In 4 of the 5 trials that included only patients with previous myocardial infarction (16,24- 25,32), myocardial infarction was recent (<4 weeks) according to current guidelines (43). In the remaining trial (18), the time interval from myocardial infarction was on average >8 weeks. When calculation regarding all-cause death was confined to the 4 trials enrolling patients with recent myocardial infarction, PCI was associated with an OR of 0.65 (95% CI: 0.37 to 1.12) from the random effects model and 0.64 (95% CI: 0.40 to 1.05) from the fixed effects model. When calculation was confined to the remaining 13 trials, PCI was associated with an OR of 0.83 (95% CI: 0.65 to 1.04) from the random effects model and 0.83 (95% CI: 0.69 to 1.00) from the fixed effects model. These results are shown in (Figure 4).(Figure 4) also shows the results of 2 additional subset analyses regarding treatment effect on mortality. The superiority of PCI was confined to the 14 trials in which coronary angiography was required before randomization (14- 21,23,26,28- 30,32). In addition, the exclusion of the 4 trials in which CABG was allowed as a treatment option in the PCI-based group (17,24- 25,27) did not make any difference in treatment effect regarding all-cause mortality: PCI was associated with an OR of 0.80 (95% CI: 0.64 to 0.99) from the random effects model and 0.80 (95% CI: 0.66 to 0.97) from the fixed effects model.

The cardiac cause of death was reported in 13 trials (14- 16,18,20- 21,23,25,27- 29,32,44). In the PCI group, 115 patients died of cardiac causes compared with 151 patients in the medical treatment group. Allocation to the PCI group was associated with a 26% reduction in the OR of cardiac death (Figure 5). There was a slight inconsistency across the trials (I² = 29%). For the same trials, we calculated the ORs for noncardiac death. A PCI was associated with an OR of 0.96 (95% CI: 0.67 to 1.36) from the random effects model and 0.95 (95% CI: 0.73 to 1.24) from the fixed effects model.

In the PCI group, 319 patients had nonfatal myocardial infarction after randomization compared with 357 patients in the medical treatment group. Allocation to the PCI group was associated with a slight and nonsignificant 10% reduction in the OR of nonfatal myocardial infarction (Figure 6). However, there was a moderate inconsistency of treatment effects across the trials (I² = 56%). A significant interaction with treatment effect regarding nonfatal myocardial infarction was observed for the proportion of patients with previous myocardial infarction (p = 0.004: the higher the proportion, the lower the OR reduction by PCI vs. medical treatment). None of the other covariates—year of completion of patient enrollment (p = 0.923), total sample size of the trial (p = 0.497), proportion of patients in the PCI group who received stents (p = 0.596) or CABG (p = 0.362), and proportion of patients in the medical treatment group who received nonprotocol revascularization (p = 0.183)—showed a significant interaction with treatment effect. We also calculated the ORs for the composite of all-cause death or nonfatal myocardial infarction. A PCI was associated with an OR of 0.80 (95% CI: 0.60 to 1.05) from the random effects model and 0.84 (95% CI: 0.74 to 0.96) from the fixed effects model.

We pooled together the results from 17 randomized trials on the value of a PCI-based treatment strategy in 7,513 patients with stable coronary artery disease. Treatment effect size was calculated from both random and fixed effects models, knowing that the former is more appropriate in case of significant heterogeneity across trials but may inflate the importance of small randomized studies. The primary analysis focused on mortality within an average follow-up period of 51 months. The analysis of the 606 death cases observed in the entire population showed that a PCI-based treatment strategy is associated with a 20% reduction in the OR of death when compared with a medical treatment-only strategy.

The effect size was not statistically dependent on any specific trial characteristics, including sample size and length of follow-up. However, it seemed to be greater among in-patients with a recent (<4 weeks) myocardial infarction, in whom there was an approximate 35% reduction in the odds of death with PCI compared with patients without recent myocardial infarction, in whom the odds of death with PCI was reduced by approximately 17%. In fact, even in a subset analysis of a recent trial, PCI was more effective than medical therapy at relieving ischemia, and rates of death and myocardial infarction were reduced in patients with absent or minimal ischemia (45). It is important to note that the included randomized trials, and consequently the entire present meta-analysis, should not be considered as a head-to-head comparison of 2 mutually exclusive treatment strategies. On the contrary, all of them evaluated the value of the PCI-based strategy as an addition to medical therapy, because patients in both study arms received medical treatment. Furthermore, 28% of the patients assigned to medical treatment only did receive revascularization at some point in time during follow-up. This might have blunted differences in survival between the 2 treatment groups in a way that cannot be predicted. Finally, individual patient data were not available for this study, which precluded several subgroup analyses. In particular, we were not able to assess the influence of the severity of ischemia at baseline on the potential benefit with PCI.

Another aspect that should be considered is that enrollment of patients was extended over a 17-year period, time in which major developments have been recorded in both the pharmacological and the interventional treatment of coronary artery disease. In fact, the year of completion of patient enrollment did not have a significant impact on the overall result as shown by the meta-regression analysis. Obviously, patients of both study arms have benefited from advances in drug therapy. Bare-metal stents were used in less than one-half of the patients included in the present meta-analysis, and drug-eluting stents were implanted in an irrelevant number of patients. Although no advantage in survival has been attributed to both bare-metal and drug-eluting stents (46- 47), this does not exclude that future advances in both pharmacological and interventional treatment of patients with coronary artery disease may reduce or further accentuate the difference in mortality observed in this meta-analysis.

Well-defined inclusion criteria relevant to the question to be addressed and a comprehensive accounting for all studies meeting those criteria are crucial to the success of the meta-analysis. The present meta-analysis intended to include all studies that investigated the relative merits of PCI in patients with stable coronary artery disease and symptoms or signs of ischemia. Two randomized trials have evaluated the role of PCI patients with persistent occluded infarct-related vessels after myocardial infarction, mostly without any clues of ischemia (22,31). In essence, this was done with the objective of assessing the open artery hypothesis. The present meta-analysis did not include the latter 2 trials because of the absence of an ischemia criterion. The same was done by Katritsis and Ioannidis in their initial (12) and updated (13) version of the meta-analysis. On the other side, we included all 4 trials that enrolled stable post-recent infarct patients with symptoms or signs of ischemia (16,24- 25,32). Only 3 of these trials (16,25,32) were included in the meta-analysis by Katritsis and Ioannidis (13). We included all 4 trials that used CABG instead of PCI in 20% to 41% of the patients in the PCI-based strategy group (17,24- 25,27). In these trials, the form of invasive treatment therapy was selected after randomization, and their exclusion from our meta-analysis would have violated the intention-to-treat principle. Katritsis and Ioannidis (13) included only 1 of these studies (25) in their meta-analysis. Thus, the present study constitutes a consistent and comprehensive investigation of available evidence by meta-analytical methods.

There is little doubt that PCI relieves ischemia and improves the exercise capacity of patients with angina pectoris (48). The results of this meta-analysis add significantly to the value of the PCI-based strategy because they contain the novel finding of a substantial reduction of long-term mortality by the use of this strategy. A reduction of similar magnitude (20%) in patients with stable coronary artery disease has not been a frequent finding of clinical trials in the past. Meta-analyses pooling the results of randomized trials on secondary prevention in patients with coronary artery disease have shown a reduction of all-cause death of 16% with statins (49) and 23% with beta-blockers (50). Nevertheless, the marginal significance level achieved in the present analysis of mortality may still require the performance of a large mortality trial to confirm the potential superiority of the PCI strategy in patients with stable coronary artery disease.

A PCI of significant coronary artery stenoses may reduce the risk of death by improving regional blood flow and stabilizing the underlying plaque through neointima formation that occurs after arterial wall injury. The risk of cardiac death was also reduced in our pooled analysis of the 13 trials that reported this event, although the lack of the information in 4 trials might have reduced the significance of this finding. The risk of nonfatal myocardial infarction was only slightly decreased. Interestingly, more recent trials showed a greater reduction in the risk of nonfatal myocardial infarction, probably because of improvements in the PCI technology and experience as well as in adjunct antithrombotic therapy. Whereas there was consistency across trials in the treatment effect size regarding all-cause mortality, there was a slight inconsistency with respect to cardiac death and a moderate inconsistency regarding myocardial infarction. This is a new clue of both the difficulties arising from pooling together results that partly depend on event definition and the robustness of all-cause mortality as an end point in the evaluation of treatment strategies. However, although the slight decrease in the risk of nonfatal myocardial infarction should not be overstated, even the lack of an increased risk of this adverse event in the PCI group may be considered a positive finding when combined with the reduced overall mortality observed with this strategy. Apparently, the PCI-based strategy is associated with a reduced risk of large myocardial infarctions leading to cardiac death, and, at least no increase in the long-term risk of smaller, nonfatal myocardial infarctions despite the known finding of myocardial injury that some patients incur early after the procedure.

These findings suggest that a PCI-based invasive strategy may improve long-term survival compared with a medical treatment-only strategy in patients with stable coronary artery disease. This justifies the performance of a new randomized clinical trial sufficiently powered for evaluating the impact of PCI on long-term mortality.