Author + information
- Received August 23, 2007
- Revision received November 21, 2007
- Accepted November 26, 2007
- Published online March 4, 2008.
- Antonio Abbate, MD⁎,⁎ (, )
- Giuseppe G.L. Biondi-Zoccai, MD†,
- Darryn L. Appleton, MBChB⁎,
- Paul Erne, MD‡,
- Andreas W. Schoenenberger, MD§,
- Michael J. Lipinski, MD∥,
- Pierfrancesco Agostoni, MD¶,
- Imad Sheiban, MD† and
- George W. Vetrovec, MD, FACC⁎
- ↵⁎Reprint requests and correspondence:
Dr. Antonio Abbate, VCU Pauley Heart Center, Virginia Commonwealth University Medical Center, 1200 East Broad Street, P.O. Box 980281, Richmond, Virginia 23298.
Objectives Our purpose was to perform a systematic review and meta-analysis of randomized trials comparing percutaneous coronary intervention (PCI) of the infarct-related artery (IRA) with medical therapy in patients randomized >12 h after acute myocardial infarction (AMI).
Background There is ongoing uncertainty about the risk–benefit ratio of late PCI in stable patients with AMI.
Methods PubMed, CENTRAL, and other databases were searched (July 2007). Studies were included if they compared PCI with medical management and randomized patients >12 h and up to 60 days after AMI, and were excluded if patients were hemodynamically unstable. Odds ratios (ORs) were pooled for dichotomous outcomes, with all-cause mortality as the primary end point. Left cardiac remodeling parameters were also pooled with generic inverse-variance weighting.
Results We retrieved 10 studies that enrolled 3,560 patients, with median time from AMI to randomization of 12 days (range 1 to 26 days), and follow-up of 2.8 years (42 days to 10 years). Randomization allocated 1,779 subjects to PCI and 1,781 to medical treatment. There were 112 (6.3%) and 149 (8.4%) deaths in the 2 groups, respectively, yielding significantly improved survival in the PCI group (OR 0.49 [95% confidence interval (CI) 0.26 to 0.94], p = 0.030). These benefits were associated with similarly favorable effects on cardiac remodeling, such as improved left ventricular ejection fraction in the PCI group (+4.4% change [95% CI 1.1 to 7.6], p = 0.009).
Conclusions Percutaneous coronary intervention of the IRA performed late (12 h to 60 days) after AMI is associated with significant improvements in cardiac function and survival.
The primary therapeutic goal in acute myocardial infarction (AMI) is to achieve early myocardial reperfusion (1). Current practice guidelines emphasize the importance of achieving reperfusion of the infarct-related artery (IRA) as early as possible (1). The reperfusion paradigm postulates indeed that early reperfusion (<12 h) results in salvage of myocardium at risk with preservation of both regional and global left ventricular function, with ensuing improvements in long-term survival (1–3). The benefit of late (>12 h) reperfusion is, however, controversial. Observational and experimental data suggest that achieving reperfusion of the IRA may confer benefits that are necrosis-independent and partially time-independent (2–4). Revascularization of the IRA, therefore, is thought to potentially prevent unfavorable post-infarction remodeling and long-term unfavorable outcomes (1–4). Smaller randomized controlled trials (RCTs) have shown mostly nonsignificant or contradictory results, and 2 recently published larger trials produced contradictory results (5,6).
The aim of our study was to perform a systematic review and meta-analysis of all RCTs comparing late percutaneous coronary intervention (PCI) of the IRA to conservative optimal medical management only in the setting of stable patients late (>12 h) in the course of AMI.
Three trained investigators (A.A., G.B.Z., D.A.) independently searched PubMed, CENTRAL, mRCT, BioMedCentral, Cardiosource, clinicaltrials.gov, and ISI Web of Science using highly sensitive and specific strategies (7). Search keywords are available in the Appendix. Searches were updated to July 2007. Proceedings from the 3 major cardiology meetings for the past 3 years were also searched. Finally, international experts were queried for additional trials.
Citations initially selected by systematic search were first retrieved as a title and/or abstract and preliminarily screened. Potentially relevant reports were then retrieved as complete manuscripts and assessed for compliance to inclusion and exclusion criteria.
Inclusion criteria were: 1) randomized allocation; 2) comparison of late PCI versus conservative medical management; 3) hemodynamic stability of patients; and 4) angioplasty performed late (>12 h after onset of symptoms) in the course of AMI.
Exclusion criteria included: 1) enrollment of patients with hemodynamic or clinical instability (e.g., cardiogenic shock), or other indications to urgent catheterization; 2) duplicate publication; 3) enrollment of patients randomized to PCI within 12 h of symptom onset; or 4) ongoing studies.
Study quality was evaluated according to the established methods of the Cochrane Collaboration. Specifically, we separately estimated the risk of selection, performance, detection, attrition bias, and allocation concealment.
The clinical outcomes of interest were defined as death, myocardial infarction, repeat revascularization, or rehospitalization for angina or heart failure (HF) at longest follow-up available, while the impact on post-infarction remodeling was assessed by means of changes in left ventricular ejection fraction (LVEF), left ventricular end-systolic volume index (LVESVI), and left ventricular end-diastolic volume index (LVEDVI). Measurements were directly abstracted when reported. Otherwise, primary authors were repeatedly contacted to obtain the precise data. Eventually, in case of no or incomplete data, these were imputed using means and standard deviations from baseline and follow-up data according to an unbiased algorithm (7).
All-cause mortality was considered the primary end point. Secondary end points were: 1) death or nonfatal AMI; 2) death, nonfatal AMI, or rehospitalization for angina or symptoms of heart failure; 3) nonfatal AMI; and 4) changes in LVEF, LVESVI, and LVEDVI.
Statistical analysis was performed using the Review Manager 4.2.4 freeware package (Plone Foundation, Houston, Texas). Dichotomous variables are reported as proportions (percentages), continuous variables as mean (standard deviation) or median (range). I2 was calculated as a measure of statistical heterogeneity, with I2 values of 25%, 50%, and 75% representing mild, moderate, and severe inconsistency, respectively. Binary outcomes from individual studies were combined with the DerSimonian-Laird random-effect model, leading to the computation of odds ratios (ORs) with 95% confidence interval (CI). Continuous variables were pooled using a random-effect inverse-variance weighting method. The number needed to treat was computed using random-effect risk differences (RDs).
In order to assess the sensitivity of the results, several subgroup analyses were performed. Specifically, we performed analyses for the primary end point according to a fixed-effect method, computing relative risks (RRs) and RDs. In addition, a funnel plot of treatment effect versus study precision was created for the primary outcome, and the risk of small study bias was appraised with the Peters’ test.
Finally, we performed meta-regression using inverse-variance-weighting (which weighs each study on the basis of its precision) to appraise the impact on death and changes in LVEF of time to PCI in the intervention group, follow-up length, total IRA occlusion as an inclusion criterion, and evidence of ischemia as an exclusion criterion.
Reported values were 2-tailed, and results were considered statistically significant at the 0.05 level. Unadjusted p values are reported throughout.
Search results and study selection
Overall, the search permitted the retrieval of 4,432 citations. We identified 20 eligible citations, which were assessed for compliance to the inclusion and exclusion criteria. We further excluded 10 studies: 6 because enrollment was allowed earlier than 12 h (8–13), 1 because it was a nonrandomized retrospective analysis (14), and 2 because the randomization process was limited to angiography and not specifically to intervention (15,16). We also excluded the TOSCA (Total Occlusion Study of Canada)-2 study (17) for the clinical outcome data as it represented duplicate data from the OAT (Occluded Artery Trial) study (5), but included the functional data for the cardiac remodeling analysis. Ultimately, 10 published studies were selected (5,6,18–25) (Fig. 1).
Baseline and follow-up LVEF data were available in 7 of the 10 studies (6,17,19,20,23,24), while no data were available from the OAT study (5) and ALKK (Arbeitsgemeinschaft Leitende Kardiologische Krankenhausärzte) study (21). Exclusion of the data deriving from the TOAT (The Occluded Artery Trial) study (25) for the cardiac function and remodeling analysis deserves specific discussion: patients enrolled in the TOAT, indeed, underwent a first assessment of cardiac function as late as 6 weeks after AMI and a second assessment at 12 months; thus, true baseline values were not available, even after contacting the primary investigator of the study.
All studies reported blinding of physicians participating in the study to randomization with regard to LVEF assessment except 1 study in which blinding was not specifically reported (20). Baseline and follow-up LVESVI and LVEDVI were available in only 4 studies (6,17,20,24).
The 10 studies included in the final analysis randomized 3,560 patients: 1,779 randomized to PCI and 1,781 to medical therapy only (Tables 1 and 2).⇓ Seven studies were multicenter trials (5,6,18,19,21–23); the remaining were single-center trials (20,23,24). The mean age was 59 years. On average, men accounted for 77% of subjects. The average of median times from AMI to intervention was 12 days (range 1 to 26 days), with the BRAVE-2 (Beyond 12 Hours Reperfusion Alternative Evaluation Trial) study (18) being the only study including only patients as early as 12 h (up to 48 h) after symptom onset, and the ALKK study (21) enrolling patients no sooner than 8 days after AMI. Angiographic documentation of total IRA occlusion was required as an inclusion criterion in 5 studies (5,19,20,24,25). Of the remaining studies, 3 required either total IRA occlusion or significant stenosis (6,21,22), while 2 studies did not define degree of IRA stenosis as inclusion criteria, but did require definitive evidence of AMI (18,23). Follow-up duration ranged from 42 days (18) to 10.2 years (6) (average for the median follow-up was 2.8 years). The average baseline LVEF was 49%. Demonstration of ischemia was necessary in all patients in 2 studies (6,18), moderate-to-severe ischemia was an exclusion criterion in 6 studies (5,19,21,23,25), whereas the remaining 2 had no specific indications regarding ischemia (22,24). Angiographic PCI success ranged from 72% (23) to 100% (20) (average of 88%). Intracoronary stents were variably used (range 0% to 100%) and glycoprotein IIb/IIIa inhibitors were not routinely used in all but 1 study (5). The medical regimen varied across the different studies and was incompletely reported in many (Table 3).
Quantitative data synthesis: clinical outcome
Considering the primary end point, late PCI for the IRA was associated with a significantly better long-term survival versus medical therapy alone (OR for death = 0.49 [95% CI 0.26 to 0.94], p = 0.030, I2 = 59%) (Fig. 1A). There were 112 deaths in the intervention arm (6.3%) versus 149 deaths in medical therapy alone arm (8.4%), yielding a number need to treat of 28 (95% CI 15 to 200). Indeed, all but 2 studies showed an OR for death favoring intervention (5,13). Similar results were obtained with random-effect estimates of RR (RR = 0.52 [95% CI 0.29 to 0.95], p = 0.028, I2 = 57%) or RD (RD = −4.3% [−7.1%; −0.2%], p = 0.032, I2 = 65%).
When compared with medical therapy alone, late intervention was also associated with concordant, albeit nonsignificant, reductions in the composite end points of death or AMI (OR = 0.70 [95% CI 0.40 to 1.23], p = 0.22, I2 = 60%) (Fig. 1B), and death, recurrent AMI, or rehospitalization for angina/heart failure (OR = 0.66 [95% CI 0.40 to 1.11], p = 0.12, I2 = 58%) (Fig. 1C). Recurrent AMI occurred with equal frequency in both groups (OR = 0.86 [95% CI 0.38 to 1.95], p = 0.72, I2 = 74%) (Fig. 1D).
Quantitative data synthesis: cardiac function and remodeling
Late PCI was associated with a significantly greater improvement in LVEF over time versus medical therapy alone, with an LVEF change of +4.4% (95% CI 1.1 to 7.6, p = 0.009, I2 = 95%) (Fig. 2). With regard to changes in LVESVI and LVEDVI, only 4 studies presented both baseline and follow-up data. Nonetheless, the difference in LVEDVI changes between the PCI group and control group was −7.0 ml/m2 (95% CI −12.18 to −1.81, p = 0.008, I2 = 89%), significantly in favor of PCI (Fig. 2). Similarly, the difference in changes in LVESVI between the PCI group and control group was −7.5 ml/m2 (95% CI −12.61 to −2.47, p = 0.004, I2 = 89%) also in favor of PCI (Fig. 2).
Sensitivity analyses for the primary end point showed results similar in direction and statistical significance for computations performed using a fixed-effect model (OR for death 0.74 [95% CI 0.57 to 0.95], p = 0.020, I2 = 64%), RRs (RR 0.52 [95% CI 0.29 to 0.95], p = 0.030, I2 = 57%), or RDs (RD −0.036 [95% CI −0.068 to −0.005], p = 0.030, I2 = 65%). Thus, these data support our overall findings, despite the evident extent of clinical and statistical heterogeneity. Indeed, point estimates and 95% CIs for ORs remained all <1.0 when sequentially excluding 1 of 7 studies (5,18–20,22,23,25), while point estimates were <1.0 but upper limits of 95% CI were >1.0 when sequentially excluding ALKK (21), Horie et al. (24), or SWISSI II (Swiss Interventional Study on Silent Ischemia Type II) (6) studies. Peters’ test did not disclose the presence of small study bias (p = 0.90), as was also demonstrated by the funnel plot (Fig. 3).
Meta-regression analysis disclosed significant interactions between several variables of interest and outcome; however, these results should be viewed as hypothesis-generating only, and lack of statistical significance should not be interpreted as lack of effect, given the well-known limitations in statistical power of meta-regression. No significant interaction between time to PCI and effect was noted. Length of follow-up showed conversely a significant interaction with changes in LVEF (p = 0.001, Beta 0.09) (Fig. 4) and a nonsignificant interaction with incidence of death (p = 0.072) suggesting that the longer the follow-up the greater the benefits observed with PCI. Studies that enrolled also patients with subtotal IRA occlusion showed a greater benefit from PCI in terms of reduction of mortality (p = 0.015), but not in terms of LVEF (p = 0.13). Trials including patients with inducible ischemia or that did not consider evidence of ischemia as an exclusion criterion showed a significantly greater benefit of PCI on LVEF (p = 0.027) and a nonsignificantly favorable impact on survival (p = 0.061) versus those trials that excluded patients with moderate-to-severe ischemia. No interaction between medications given at discharge and benefits from PCI was found (all p > 0.10).
Table 4 shows design features for the appraisal of the internal validity of included studies. No association between design features and effects was observed.
This systematic review is based on the statistical pooling of 10 RCTs that enrolled in total more than 3,500 patients comparing PCI of the IRA with conservative medical management in patients presenting more than 12 h after an AMI. Notwithstanding the limitations inherent to the primary studies, this meta-analysis includes the largest cohort to date and shows a statistically significant survival and remodeling advantage in favor of PCI of the IRA late after AMI.
Significant findings and clinical implications
The findings of a beneficial effect of late PCI proposes that revascularization of the IRA by means of PCI occurring late and beyond the window of myocardial salvage can be favorable in terms of post-AMI cardiac remodeling and all-cause mortality. The clinical implications of such findings are potentially very significant. The number of patients treated within 12 h of the onset of symptoms is still disappointing, with 8.5% to 40% of patients presenting beyond 12 h (18).
The “open artery hypothesis” speculates that restoration of anterograde blood flow to the peri-infarct area is beneficial to the myocardium even late and beyond the time limit set for salvage from myocardial necrosis (2–4). A detailed discussion of the proposed mechanism by which late reperfusion is associated with late improvement in left ventricular function is beyond the scope of this review. Salvage of myocardium at risk (“hibernating” myocardium) from death due to apoptosis may be a plausible explanation (4,26,27). Ischemia can indeed induce apoptosis in cardiomyocytes and even late reperfusion may favorably affect the apoptotic cascade (4,26,27). Moreover, ischemia itself stimulates formation of collateral circulation, which in the setting of AMI can serve to preserve some degree of retrograde perfusion, potentially extending the period of viability of myocardium at risk (28). Viability of stunned myocardium has indeed been demonstrated weeks after AMI, suggesting that reperfusion of the IRA could interrupt the progression from hibernating myocardium to necrotic/apoptotic myocardium (29).
It is worth mentioning that while the TOAT study (25) demonstrated a trend towards adverse remodeling in the PCI arm, a substudy on 26 patients who underwent dobutamine stress cardiac magnetic resonance imaging to assess myocardial viability demonstrated that PCI had a beneficial effect on remodeling in patients with demonstrable myocardial viability (30).
According to these premises, late PCI of the IRA was associated with preservation of left ventricular function and more favorable cardiac remodeling in our data. These data suggest that late PCI may prevent death by preventing the occurrence of ischemic cardiomyopathy. The lack of difference in terms of recurrent AMI suggests that prevention of recurrent AMI is not necessarily the means by which late PCI prevents death.
Validity and potential limitations of the meta-analysis
As with any meta-analysis, there are inherent difficulties in using data from multiple studies of a similar nature to derive conclusions about an overall effect of a particular intervention. Substantial statistical inconsistency was detected in this study, which highlights the presence of significant differences between the included studies, without, however, questioning the validity of the statistical analysis performed using a random effect pooling. The 2 most recently published studies in this analysis (OAT  and SWISSI II ) have important differences in study design and length of follow-up and, not surprisingly, show conflicting results; SWISSI-II (6) randomized only patients with proven silent ischemia on stress testing, whereas the OAT (5) excluded those patients with post-infarction angina and/or moderate-to-severe ischemia. Our analysis suggests, indeed, that studies that did not exclude patients with ischemia showed a greater benefit from late PCI. However, one-third to two-thirds of patients have residual symptomatic or silent ischemia after AMI (31,32), and there is some debate on whether the OAT study is representative of real life treatment scenarios (33). Recruitment in the OAT study was indeed difficult and interrupted early: only 2 patients per year per center were enrolled on average representing a very small minority of patients (5,33). In addition, length of follow-up was dramatically different between these 2 studies, with the SWISSI II following patients for a median of 10.2 years, while less than one-half of the patients in the OAT study had a follow-up that reached 3 years. Moreover, survival curves in the OAT study at 4 years were constructed using estimates, which may have overestimated event rates in the intervention arm due to early attrition of procedural-related events and underestimated late effects of intervention (i.e., prevention of heart failure), which may become apparent only several years later (5,33). Of note, only 3 individual studies have shown a significant improvement in death or death and heart failure (ALKK , Horie et al. , and SWISSI II ), and all 3 studies had a median follow-up of >4 years.
Moreover, many studies, including the OAT (5), required that all patients had demonstrated total occlusion of the IRA in order to qualify for randomization, whereas others, including SWISSI II (6) and ALKK (21), required the presence of a hemodynamically significant IRA stenosis for inclusion, thus including also patients with subtotally occluded IRA. Variable severity of the IRA lesion could certainly affect the size of the benefit observed from PCI, with greater benefits for subtotal occlusions; although the great majority of patients included in the studies (84%) had a totally occluded IRA, our analysis suggests greater benefits when patients with subtotal IRA occlusions were also included.
Implications for further research
The available evidence from individual trials examining the effects of late PCI has not been able to establish a clear benefit. The data from our meta-analysis suggests that a survival benefit may indeed exist. Patient selection and adequate length of follow-up may prove to be essential in determining who is likely to benefit the most from late PCI. Patients symptomatic for angina or heart failure and those with residual ischemia or documented viability are more likely to benefit from late PCI at a long-term follow up. Patients with uncomplicated AMI, especially if with reduced life expectancy, however, may not benefit from routine PCI. Even in the absence of significant effects on hard end points, the OAT study showed that patients treated with PCI were significantly less likely to be symptomatic for angina at 4, 12, and 24 months (5). The OAT study (5), however, also showed a potential for harm associated with late PCI, in terms of an increased rate of recurrent AMI. This has not been confirmed in our meta-analysis and did not affect survival in the OAT cohort, but it illustrates the inherent limitations and potential long-term complications of PCI as a means of achieving sustained reperfusion, and highlights the need for further refinement in PCI technology and adjunctive medical therapy to improve longevity of IRA patency.
This review presents a meta-analysis of data from 10 RCTs in which patients treated late (>12 h) during the course of AMI were randomized to PCI of the IRA versus conservative management and shows a survival advantage in performing PCI late in the course of an AMI, associated with a more favorable post-infarction cardiac remodeling pattern.
The authors wish to thank Dr. Vera Di Trocchio (Virginia Commonwealth University, Richmond, Virginia) for her editorial assistance with the manuscript.
Appendix Search Strategy
The investigators independently searched PubMed, CENTRAL, mRCT, BioMedCentral, Cardiosource, clinicaltrials.gov, and ISI Web of Science using highly sensitive and specific strategies (7). Search keywords included ‘randomized,’ ‘percutaneous coronary intervention,’ ‘PCI,’ ‘stent,’ ‘angioplasty,’ ‘revasc*,’ ‘recanaliz*,’ ‘acute myocardial infarction,’ ‘AMI,’ ‘infarct*,’ ‘occlusion,’ and ‘occlu*’ (where * denotes a wildcard). Searches were updated to July 2007. No language restriction was used. Proceedings from the annual American Heart Association, American College of Cardiology, and European Society of Cardiology meetings for the past 3 years were also manually searched. Finally, international experts were queried for additional trials.
Dr. Agostoni has received lecture fees from Cordis Belgium and Jolife Inc.; Dr. Biondi-Zoccai has consulted for Boston Scientific, Cordis, and Mediolanum Cardio Research, and has received lecture fees from Bristol-Myers Squibb; Dr. Vetrovec was on the December 2006 FDA panel for drug-eluting stents and has received various research and/or teaching grants, honoraria, consulting fees and/or stock in the following corporations over the past 5 years: Pfizer, Cordis/Johnson&Johnson, CV Therapeutics, Boston Scientific, Medtronic, Schering-Plough, and Abbott Labs; it should be noted that the U.S. FDA did not identify a substantial conflict of interest (to preclude his participation) to the December 2006 Panel Meeting for drug-eluting stents. This study is part of a training project of the Meta-analysis and Evidence-based Training in Cardiology (METCARDIO) Center, Richmond, Virginia. Drs. Abbate, Biondi-Zoccai, and Appleton contributed equally to this work.
- Abbreviations and Acronyms
- acute myocardial infarction
- confidence interval
- infarct-related artery
- left ventricular ejection fraction
- left ventricular end-diastolic volume index
- left ventricular end-systolic volume index
- odds ratio
- percutaneous coronary intervention
- randomized controlled trial
- risk difference
- relative risk
- Received August 23, 2007.
- Revision received November 21, 2007.
- Accepted November 26, 2007.
- American College of Cardiology Foundation
- Antman E.M.,
- Anbe D.T.,
- Armstrong P.W.,
- et al.
- Braunwald E.
- Kim C.B.,
- Braunwald E.
- Lipinski M.J.,
- Biondi-Zoccai G.G.L.,
- Abbate A.,
- et al.
- Ellis S.G.,
- da Silva E.R.,
- Heyndrickx G.,
- et al.
- Vermeer F.,
- Oude Ophuis A.J.,
- vd Berg E.J.,
- et al.
- Sutton A.G.,
- Campbell P.G.,
- Graham R.,
- et al.
- Topol E.J.,
- Califf R.M.,
- Vandormael M.,
- et al.
- van Loon R.B.,
- Veen G.,
- Kamp O.,
- et al.
- Dzavik V.,
- Buller C.E.,
- Lamas G.A.,
- et al.,
- TOSCA-2 Investigators
- Steg P.G.,
- Thuaire C.,
- Himbert D.,
- et al.
- Silva J.C.,
- Rochitte C.E.,
- Junior J.S.,
- et al.
- Zeymer U.,
- Uebis R.,
- Vogt A.,
- et al.
- Ellis S.G.,
- Mooney M.R.,
- George B.S.,
- et al.
- Dzavik V.,
- Beanlands D.S.,
- Davies R.F.,
- et al.
- Horie H.,
- Takahashi M.,
- Minai K.,
- et al.
- Yousef Z.R.,
- Redwood S.R.,
- Bucknall C.A.,
- Sulke A.N.,
- Marber M.S.
- Abbate A.,
- Bussani R.,
- Biondi-Zoccai G.G.,
- et al.
- Abbate A.,
- Bussani R.,
- Biondi-Zoccai G.G.,
- et al.
- Brunken R.,
- Tillisch J.,
- Schwaiger M.,
- et al.
- Bellenger N.G.,
- Yousef Z.,
- Rajappan K.,
- Marber M.S.,
- Pennell D.J.
- Giannuzzi P.,
- Marcassa C.,
- Temporelli P.L.,
- et al.
- Coletta C.,
- Sestili A.,
- Seccareccia F.,
- et al.
- Rutherford B.D.