Risk of Stroke in Patients with Stable Coronary Artery Disease Undergoing Percutaneous Coronary Intervention versus Optimal Medical Therapy: Systematic Review and Meta-Analysis of Randomized Controlled Trials

Background Stroke is a rare but serious adverse event associated with percutaneous coronary intervention (PCI). However, the relative risk of stroke between stable patients undergoing a direct PCI strategy and those undergoing an initial optimal medical therapy (OMT) strategy has not been established yet. This study sought to investigate if, in patients with stable coronary artery disease (SCAD), an initial strategy PCI is associated with a higher risk of stroke than a strategy based on OMT alone. Methods We performed a meta-analysis of 6 contemporary randomized control trials in which 5673 patients with SCAD were randomized to initial PCI or OMT. Only trials with stent utilization more than 50% were included. Study endpoint was the rate of stroke during follow up. Results Mean age of patients ranged from 60 to 65 years and stent utilization ranged from 72% to 100%. Rate of stroke was 2.0% at a weighted mean follow up of 55.3 months. On pooled analysis, the risk of stroke was similar between patients undergoing a PCI plus OMT and those receiving only OMT (2.2% vs. 1.8%, OR on fixed effect = 1.24 95%CI: 0.85–1.79). There was no heterogeneity among the studies (I2 = 0.0%, P = 0.15). On sensitivity analysis after removing each individual study the pooled effect estimate remains unchanged. Conclusions In patients with SCAD an initial strategy based on a direct PCI is not associated with an increased risk of stroke during long-term follow up compared to an initial strategy based on OMT alone.


Introduction
In patients with stable coronary artery disease (SCAD) the role of percutaneous coronary intervention (PCI) is controversial. Indeed, data from recent randomized clinical trials (RCT) [1][2][3] and meta-analyses [4][5][6] have consistently shown that in patients with SCAD, PCI does not reduce the risk of death and myocardial infarction (MI) compared with optimal medical treatment (OMT). Furthermore the attempts to identify patients at higher risk, such as those with documented myocardial ischemia [7] or diabetes [2] that would benefit from an interventional strategy have failed. On the other hand, in most of the randomized comparisons, the patients with SCAD treated with PCI had a better and faster symptoms relief than those treated medically. Yet, a substantial proportion of patients belonging to the latter group crossed over and underwent PCI. Thus, the choice of referring patients to an initial conservative strategy based on either the optimization of medical treatment or a direct PCI strategy plus OMT should take into account several factors such as the presence and the intensity of symptoms, the patient's risk profile (for example patients with marked signs of myocardial ischemia, systolic left ventricle dysfunction, high risk coronary obstructions, comorbidities that have been scarcely represented in randomized clinical trials) and intervention hazards. Regarding this latter aspect, stroke following PCI represents undoubtedly a dismal complication with severe disability and high mortality [8,9]. Data from large registries [8,9] suggest that guiding catheter manipulation, use of anticoagulant and potent antiplatelet inhibitors (such as glycoprotein IIb/IIIa), especially in technically difficult PCI [9], may increase the risk of ischemic and hemorrhagic stroke, respectively. However, the relative risk of stroke between stable patients undergoing a direct PCI strategy and those undergoing an initially conservative strategy has not been established yet. Indeed, all RCTs comparing PCI with OMT in stable patients were far underpowered to detect a difference in terms of such a rare event as stroke. Therefore, we performed a systematic review and meta-analysis of contemporary RCTs to investigate if PCI plus OMT is associated with a higher risk of stroke than OMT alone in patients with SCAD.

Study selection and study endpoint
We carried out a systematic review of the available publications according to the current PRISMA guidelines to perform meta-analyses of RCT [10] (S1 Table). We searched for relevant articles, filtered by RCT, published in the MEDLINE and the Cochrane Library using the following key words that were variously combined (S1 Text): "stable angina", "stable coronary artery disease", "medical therapy", "PCI", "stent". No language restriction was used. We also checked the reference lists of reviews and relevant articles.
Inclusion criteria were as follows. 1. RCT design, 2. Comparison between PCI plus OMT and OMT alone, 3. Stent implantation in more than 50% of PCI patients 4. Data on stroke available.
For studies that compared OMT with separate arms of PCI and coronary bypass only data on comparison between OMT and PCI were extracted. We excluded trials that included patients with a recent (within 6 months) episode of acute coronary syndrome. Trials that compared OMT with revascularization as a whole (either PCI or coronary bypass) were also excluded. Two investigators (N.T, G.G.) independently reviewed the titles, abstracts, and studies to determine whether they met the inclusion criteria. Conflicts between reviewers were resolved by consensus.
The study endpoint was the rate of stroke based on the definition applied in each study. For the endpoint we used data from the longest follow up available to a maximum of 5-year follow up.
Data were extracted on the basis of the intention-to-treat populations.

Statistical Analysis
Odd Ratios (OR) and 95% Confidence Intervals (95% CI) were used as the summary statistic. The pooled OR was calculated by using both fixed-effect (inverse variance weighted) and random-effect (DerSimonian and Laird) models. Between-study heterogeneity of effects was analyzed using the χ 2 and inconsistency across study results quantified by I 2 statistics, with I 2 <25%, 25% I 2 50%, and I 2 > 50%, respectively, representing mild, moderate, and severe inconsistency. Sensitivity analysis was performed by evaluating the influence of removing individual studies on the pooled OR. The possibility of publication bias was assessed both visually by funnel plot and Harbord's test. Statistical analyses were performed using Stata/SE 11.2 (StataCorp LP, College Station, TX).  Table 1 shows the main characteristics of the studies included. Stent utilization ranged from 72% to 100%. Definition of optimal medical therapy was provided in all studies but one. Most of the studies had 5-year follow up available [1,2,13,14], 2 studies reported data on a shorter follow up (1-year [12] and 2-year [3], respectively).

Results
Mean age of patients ranged from 60 to 65 years with a weighted mean age of 62.3 years. All studies enrolled predominantly male patients ( Table 2). Utilization of aspirin was equal or greater than 80% in all studies. Beta-blockers and statins were used in more than 60% of patients with the exception of 1 trial [13] in which they were used in nearly 50%. Angiotensin converting enzyme inhibitors or angiotensin receptor blockade were used in more than 60% in most studies.
Among the 5668 patients included in the main analysis [in the Japanese stable angina pectoris study (JSAP) [13] outcomes were not reported for 5 patients) the rate of stroke was 2.0% at a weighted mean follow up of 55.3 months. On pooled analysis, the risk of stroke was similar between patients undergoing PCI plus OMT and those receiving only OMT (2.2% vs. 1.8%, OR on fixed effect = 1.24 95%CI: 0.85-1.79). There was no heterogeneity among the studies (I 2 = 0.0%, P = 0.915) [Fig 2]. The funnel plot was very symmetric showing the lack of publication bias (Fig 4). The Harbord test further supports this result (p value for small-study effect = 0.765).

Discussion
The main result of the present study, drawn from 6 RCTs including 5673 patients with SCAD followed up for a mean time of 55 months, is that in patients with SCAD an initial strategy of PCI plus OMT is not associated with an increased risk of stroke compared to OMT alone.
Although PCI is widely used for the management of patients with SCAD its role is controversial [15]. Indeed recent clinical trials(1-3) and meta-analyses [4][5][6][7] have shown that in patients with stable angina an initial PCI strategy plus medical therapy does not reduce the risk of death or MI compared to OMT alone. On the other hand from the same sources stem the evidences that PCI is associated with a better and more rapid symptoms relief. Accordingly whether a patients with stable angina should be referred to invasive management or not still represents a major question that physicians have to face in daily clinical practice. In particular among factors that may influence this choice the evaluation of intervention hazards should be taken into account. Stroke is a rare but devastating complication related to PCI [8,9]. Although its incidence, risk factors and outcome have been widely described [8,9,[16][17][18] a direct comparison in patients with SCAD aimed to investigate if initial PCI plus OMT strategy is associated with a higher risk of stroke than OMT alone has not been performed so far.
The results of our meta-analysis, showing that the risk of stroke is not different between patients managed with PCI and those managed with OMT, are therefore noteworthy.   The strength of this findings rely on the satisfaction of all requirements for meta-analysis in term of low heterogeneity, no publication bias and sensitivity analyses.
There may be different reasons to explain our findings. First, the risk of stroke in patients undergoing PCI has been historically associated with procedural factors with an incidence during hospitalization less than 0.4% [16,17]. Guide catheter manipulation, use of anticoagulant and potent antiplatelet inhibitors may increase the risk of ischemic and hemorrhagic stroke, respectively. In a recent study (18) from the British Cardiovascular Intervention Society, including 426046 patients undergoing PCI, the incidence of 30-day ischemic and haemorrhagic stroke was 0.10% and 0.03%. Patients with SCAD were at lower risk compared to patients with ACS.
This quite low risk could also be paid by a considerable portion of patients treated with OMT alone due to the high rate of cross over during long-term follow up, realizing a sort of catch up in the risk of stroke. Indeed, the need of elective PCI in the OMT arm ranged from 9% of The Medicine, Angioplasty, or Surgery Study II study to 36.5% of the JSAP study [in the COURAGE study and Fracional Flow Reserve versus Angiography for Multivessel Evaluation 2 (FAME II) trial the rate of elective PCI for persistent angina was 32.6% and 26.5%, respectively).
Second, the risk related to PCI could be partly outfitted during follow up by the fact that patients undergoing to PCI are more likely to receive dual antiplatelet therapy compared to patients managed medically due to stent implantation routinely used in contemporary clinical practice. In fact, it is known that dual antiplatelet therapy reduce the risk of stroke as a whole compared to aspirin alone [19,20]. Notably in the present meta-analysis in order to reflect a contemporary clinical practise we included only study with a rate of stent implantation more than 50%.

Study limitations
The results of the present study should be interpreted in light of some limitations. This meta-analysis was based on aggregate data of randomized clinical trials that traditionally are conducted in highly-experienced centres and enrol patients with low risk clinical profile. Therefore, generalization of effect estimates should be undertaken with caution. Indeed, traditional risk factors for stroke in patients undergoing PCI such as advance age, prior stroke, atrial fibrillation and procedural complexity may be less represented in clinical trials compared to a real world setting. Given these observations whether PCI is associated or not with an increased risk of stroke as compared to OMT also in higher risk patients should be evaluated by further investigations.
Although in the present study we included only trials with > 50% of stent utilization to reflect a modern PCI practise it should be taken into account that these trials were conducted in different era in terms of type of stent, type and duration of dual antiplatelet therapy (this latter aspect related to the rate of drug eluting stent utilization). In the present study we included more than 5000 patients; it is not excludible that a study with a larger sample size could be able to detect a significant higher risk of stroke in patients undergoing a PCI based approach compared to OMT alone. To date, however, there are no larger studies available. Nevertheless, assuming a stroke rate of 1.9% in the PCI group and 1.2% in the OMT group; i.e. 37 relative risk reduction [based on data from the Clinical Outcome Utilizing Revascularization and Aggressive Drug Evaluation (COURAGE) study] 5170 patients would be required for a study power of 80% with a type I error of 0.05. Finally, it would be of interest to evaluate also the early risk of stroke related to coronary interventions however data on peri-procedural stroke and timing of stroke were missing in most of the studies included in the present meta-analysis. Nonetheless data from recent meta-analyses [21] of 19 trials including 10944 patients randomized to coronary bypass versus PCI showed that the 30-day rate of stroke in the PCI arm was quite low (1.2% and 0.34, respectively). Yet, these findings have been confirmed also in a recent European survey showing in stable patients a rate of in hospital stroke following PCI of 0.3% [9]. Therefore, assuming a stroke rate of 0.3% in the PCI arm to detect a significant relative risk reduction of 37% in the OMT arm, with a study power of 80% and an alfa level of 0.05, 9920 patients should be recruited.

Conclusions
In this meta-analysis of available contemporary RCTs including patients with SCAD an initial strategy based on a direct PCI is not associated with an increased risk of stroke during longterm follow up compared to an initial strategy based on OMT alone.  Estimation of publication bias. A funnel plot of SE versus ln(OR) for the meta-analysis. When the standard error is used on the y axis of a funnel plot, it is conventional to reverse the axis so that the most precise studies are displayed at the top of the plot. They tend to cluster near the mean effect size. Smaller studies are depicted at the bottom of the graph and tend to be dispersed in a range of value. In absence of publication bias as shown in the present figure the studies are distributed symmetrically about the combined effect size. The pseudo 95% confidence limits illustrate the expected 95% confidence interval about the pooled fixed-effects estimate for the meta-analysis.