Impact of Collateral Flow to the Occluded Infarct-Related Artery on Clinical Outcomes in Patients With Recent Myocardial Infarction: A Report From the Randomized Occluded Artery Trial

Angiographically visible coronary collaterals in humans have been postulated to exert a beneficial effect on infarct size,¹ ventricular aneurysm formation,^2,3 viability, and ventricular function after acute coronary occlusion.^3–5 In acute myocardial infarction (MI) treated by primary percutaneous coronary intervention (PCI), however, there have been conflicting reports on this postulated link and observed clinical outcomes.^6,7

The randomized Occluded Artery Trial (OAT) tested routine PCI for persistently occluded infarct-related coronary arteries 3 to 28 days post-MI. It afforded an opportunity to examine the relationship between visible collateralization of the infarct-related artery (IRA) and long-term outcome in a large rigorously characterized cohort of MI survivors.⁸ Specifically, OAT mandated centrally analyzed coronary angiography at baseline (median 8 days from infarction to randomization) that included semiquantitative assessment of ipsilateral and contralateral collateral flow in both study arms (PCI or optimal medical therapy alone).⁹ Long-term adjudicated outcomes were then collected.

As in acute MI, the degree of visible collateral flow during both subacute and chronic post MI phases has the potential to influence late clinical outcome through preserved viability, altered ventricular geometry, or electric substrate. Moreover, it is conceivable that the degree of collateralization modulates the influence of late infarct artery recanalization. This interaction might operate by preserving viability such that PCI both relieves ischemia and enables regional left ventricular (LV) recovery.¹⁰ Alternatively, late recanalization of persistently occluded infarct arteries might be most beneficial in patients without the robust collateral flow implied by easily observed collateralization to the infarct artery territory.¹¹ In the present analysis, we sought to assess the association between baseline collaterals and overall clinical outcomes and to explore a potential interaction between collaterals and PCI versus medical therapy alone.

Methods

The methods used in OAT have been described in detail previously.⁹ In brief, 2201 patients were enrolled (2166 between February 2000 and December 2005 in the main OAT study, and 35 patients in the extension phase of OAT-NUC [Viability and Remodeling in OAT Ancillary Study] substudy in 2006) and observed for a mean of 3.2 years. By 5 years, 30% either had reached a primary outcome or were still being followed-up. Patients were eligible for enrollment if coronary angiography, performed on calendar days 3 to 28 (minimum 24 hours) after myocardial infarction, showed total occlusion of the IRA with poor or absent antegrade flow (Thrombolysis In Myocardial Infarction [TIMI] flow grade of 0 or 1) and if they met a criterion for increased risk (left ventricular ejection fraction ≤50%, proximal occlusion of a major epicardial vessel, or both). Patients were excluded from the study if they had New York Heart Association (NYHA) class III or IV heart failure (HF), shock, serum creatinine ≥2.5 mg/dL (221 μmol/L), angiographically significant left main or 3-vessel coronary artery disease, angina at rest, or severe ischemia on stress testing. Patients were randomly assigned to PCI with stent placement and optimal medical therapy (PCI group) or optimal medical therapy alone (MED group). Images from the baseline angiogram collected in both groups were reviewed (by an angiography core laboratory blinded to treatment assignment), and collaterals were graded in 3 categories according to a modified TIMI phase I collateral score,¹² whereby in grade 0, there is no angiographic filling of the distal vessel, in grade 1, there is faint opacification of the distal vessel only or only small segments are visualized, and in grade 2, the entire distal vessel is visualized and densely opacified. In the case of several pathways per lesion, the largest one was considered for further analysis. The primary end point of OAT (and therefore used in this secondary analysis) was a composite of death from any cause, reinfarction, and hospitalization for NYHA IV HF. Study sites submitted clinical records pertaining to all hospitalizations for cardiovascular events and pneumonia for review to avoid ascertainment bias. Whether HF was responsible for these hospitalizations was centrally confirmed according to prespecified criteria, and NYHA class at the time of hospitalization was determined by an independent event committee whose members were blinded to treatment assignment. The specific process of central adjudication of HF events has been previously described.⁹ For this report, centrally confirmed hospitalizations for both NYHA class III and NYHA class IV HF are reported. Mortality events and their cause were centrally reviewed.

Secondary end points included each component of the primary end point, class III or IV heart failure, as well as composites of each component with death. The prespecified definition of reinfarction required 2 of the following 3 criteria: symptoms for ≥30 minutes, electrocardiographic changes, and elevated cardiac markers.

Baseline characteristics and outcomes were studied in groups of patients divided as a function of baseline angiographic collaterals using conventionally defined collateral flow grade describing only the most robust collateral supply. These analyses were done in a binary fashion (ie, comparing patients without collaterals to patients with grade I or II collaterals) as well as by analyzing the best observed collateral flow by semiquantitative grade. The analyses were then repeated after applying a novel “total flow” score in which all sources of residual flow to the distal IRA were summed (total flow = sum of collateral grades [0 to 2] from all non-IRA sources + residual TIMI flow grade in IRA). In a second step, outcomes were analyzed as a function of initial randomization (ie, PCI and optimal medical therapy versus optimal medical therapy alone), with analyses for interactions.

Statistical Analysis

Baseline characteristics of study patients were summarized as frequencies and percentages for categorical variables and as the means with SD for continuous variables and compared within the collateral grade category with Mantel-Haenszel χ² test for trend or ANOVA, respectively. (The Mantel-Haenszel χ² statistic tests the alternative hypothesis that there is a linear association between the row variable and the column variable.) Kaplan-Meier curves for patients within collateral grade levels were compared with log-rank tests. Interaction between study assigned treatment and the collateral grade was assessed with use of a Cox regression model.

The degree of collaterals was modeled as an indicator variable with the hazard ratio (HR) for the estimate of the 5-year primary and secondary outcomes in patients with grade 0 collaterals compared with that of patients with grade 1 or 2 collaterals. Analyses were performed according to the intention-to-treat principle. The variables were chosen by Cox models using backward elimination with P≤0.01 to remain in the model. All 37 standard baseline variables were considered for these models. The methods are similar to those used in Kruk et al.¹²

To reduce the likelihood of type I errors, the OAT protocol required that all secondary analyses employ a P value threshold of P≤0.01.⁹ SAS version 9.1.3 (SAS Institute, Cary, NC) was used for statistical analyses.

Results

Core laboratory data describing collateral flow to the infarct artery territory were available for 1087 and 1086 (99% of all randomized patients) of the patients randomly assigned to PCI or MED, respectively. The distribution of angiographic collaterals at baseline was similar between the 2 arms: 251 patients (134 [12.3%] PCI and 117 [10.8%] MED) had no visible collaterals, whereas 1922 patients had grade I (778 [71.6%] PCI and 770 [70.9%] MED) or II (175 [16.1%] PCI and 199 [18.3%] MED) collateral support to the occluded IRA. The baseline and angiographic characteristics of the patients according to collateral grade are described in Table 1. Overall, patients with better collateral flow grades were more likely to have characteristics suggesting lower baseline risk, including younger age, freedom from prior diabetes or HF, lower heart rate, lower fasting glucose, higher left ventricular ejection fraction, and a culprit artery other than the left anterior descending artery. Well-developed collaterals were also associated with a longer interval from symptom onset to diagnostic coronary angiography (Table 1).

Collaterals and the Primary End Point

Cumulative 60-month life-table estimated event rates according to the presence or absence of collaterals are summarized in Table 2, and the relation between angiographic collateral flow grade and clinical outcomes at 60 months is depicted in Figure 1.

• Download figure
• Download PowerPoint

Compared with patients without visible collaterals, those with collaterals tended to reach the primary end point less frequently (16.9% versus 22.7%; P=0.014). This trend, which failed to reach statistical significance, was observed in both treatment arms (18.0% versus 25.4% in the PCI arm; 15.9% versus 18.7% in the MED arm). Similarly, and again regardless of randomized treatment assignment, the primary end point tended to occur less commonly as the angiographic collateral grade increased (P for trend between 0.01 and 0.05; Figure 1). Using the Cox proportional hazard model (Figure 2), the HR for the occurrence of primary end point (death, MI, or congestive heart failure [CHF] class IV) in patients who had any baseline collaterals was 0.68 (99% confidence interval [CI], 0.46 to 1.02; P=0.014) compared with patients with no visible collaterals at baseline angiography. By multivariate analyses, however (adjusting first for variables distributed unevenly by collateral grade [history of diabetes mellitus, hypertension, Killip 2 to 4 class, left anterior descending coronary as culprit artery, rales on baseline physical examination, left ventricular ejection fraction, ST-segment elevation, and age] and separately by variables shown to be predictors of the OAT primary outcome [estimated glomerular filtration rate, left ventricular ejection fraction, days elapsed from qualifying myocardial infarction to randomization, history of peripheral vessel disease, history of CHF, history of diabetes mellitus, rales on baseline physical examination], or of death [estimated glomerular filtration rate, left ventricular ejection fraction, days from qualifying MI to randomization, Killip 2 to 4 class, history of angina, history of cerebrovascular disease, history of CHF]), the presence of collaterals was found not to be an independent predictor of the primary end point (P=0.67 and P=0.85, respectively) nor of death (P=0.79 and P=0.75, respectively).

• Download figure
• Download PowerPoint

There was no interaction between the presence of collaterals and the assigned treatment (PCI versus MED) on occurrence of the primary end point both when angiographic collaterals were analyzed as present or absent (Figure 3B and Table 2) or by angiographic collateral grade (Table 3). Similar observations were made when examining event rates at various time intervals during follow-up (as opposed to cumulative 60-month event rates; online-only Data Supplement).

• Download figure
• Download PowerPoint

Collaterals and Secondary Outcomes

There was no difference in the occurrence of death or reinfarction between patients with and without collaterals overall. Similarly, there was no interaction between the presence of collaterals and the assigned treatments (Tables 2 and 3). However, when collaterals were examined by grade (Figure 1), an inverse association between collateral grade and death (P for trend=0.009) was observed. There was no such association for reinfarction.

Patients with visible collaterals had a significantly lower incidence of class III or IV HF (5.2% versus 11.6%; P<0.001) in both treatment arms (Table 2). In a Cox proportional hazards model (Figure 4), the HR for the occurrence of CHF class III or IV in patients who had any visible collaterals was 0.43 (99% CI, 0.24 to 0.76; P<0.001) compared with patients with no visible collaterals. When fatal events were included (the secondary composite of death or class III to IV heart failure), this relationship persisted (Tables 2 and 3). Importantly, when the presence of collaterals was examined after adjustment for variables distributed differently at baseline, or for variables predicting the primary outcome, neither the primary outcomes nor any of the secondary outcomes or their composites examined in Table 2 were significantly influenced by collaterals.

• Download figure
• Download PowerPoint

For class III to IV heart failure, we observed a trend toward interaction between the presence of collaterals and treatment effect such that patients without collaterals assigned PCI had a numerically lower incidence of HF when compared with similar patients assigned MED, whereas those with collaterals did not (P for interaction=0.02). No such trend was evident when the composite of class III to IV CHF or death was analyzed. Moreover, no interaction existed between collaterals and assigned treatment for the composite of death or class III to IV HF (P for interaction=0.60 when collaterals were defined in binary fashion, Table 2, or P=0.50 when collateral grade was examined, Table 3). In patients with visible collaterals, the randomized assignment to PCI or MED had no effect on the occurrence of death or CHF class III or IV, and the event curves completely overlap (HR=1.00; 99% CI, 0.71 to 1.42; P=1.0; Figure 3A). In patients without collaterals, a suggestion of an early benefit of PCI was no longer evident after 5 years of follow-up (HR=0.84; 99% CI, 0.39 to 1.81; P=0.56; Figure 3A and Table 3).

Impact of Total Flow to the IRA

In an attempt to better characterize residual baseline (pre-PCI) flow to the infarct artery, we analyzed baseline characteristics and outcome events of patients as a function of total flow, that is, the sum of collateral grades (0 to 2) from all non-IRA sources and the residual TIMI flow grade in IRA. Patients were grouped in 3 categories: no flow (grade 0 total flow grade), minimal flow (grade 1), and residual flow (grades 2 to 4; there were no patients with total flow score 5). As was the case for patients with collaterals, patients with greater total flow had consistently more favorable baseline risk characteristics, such as a lower age, a lower frequency of diabetes, previous MI, anterior location, multivessel disease, Killip class 2 to 4, S3 rales, ST-segment elevation (data not shown). When outcome events were analyzed as a function of total flow (Table 4), greater total flow was also consistently associated with better outcomes for all the outcome events analyzed (death, CHF class III or IV, the primary composite end point, and the composite of death and CHF) except fatal and nonfatal MI, which was evenly distributed across total flow groups. There was no interaction between total flow and randomized assignment to PCI or MED, except for CHF III or IV (P=0.008), but this interaction disappeared when fatal events were included (P=0.23). By multivariate analyses (adjusting for variables distributed differently by total flow grade and by risk factor variables for the primary outcome), the total flow grade was neither an independent predictor of the primary outcome (P=0.09 and P=0.10 respectively) nor of death (P=0.10 for both), or of CHF III or IV (P=0.10 and P=0.11 respectively).

Discussion

The open artery hypothesis suggests that late recanalization of an occluded IRA is associated with clinical benefit, independent of the myocardial salvage achieved by timely recanalization of an infarct vessel.¹³ This stemmed from observations that late patency of an IRA was independently associated with better LV function, electric stability, and the potential for provision of collateral vessels to other coronary beds for during future ischemic events. However, the OAT demonstrated no reduction in death, recurrent MI, or new class IV HF attributable to routine PCI applied to persistently occluded IRAs in stable patients with persistent IRA occlusion documented 3 to 28 days after MI.⁸ Results from the OAT, supported by those from smaller previous trials¹⁴ have now been incorporated into both US¹⁵ and European¹⁶ guidelines for the management of patients with acute MI.

For the same reasons, collaterals in patients with ischemic heart disease have generally been viewed as protective against the development of HF, reinfarction, or death.^17,18 Our study confirms that angiographically visible collaterals to the IRA territory at baseline angiography post-MI are indeed associated with a lower risk of heart failure. Moreover, increasing collateral grade was associated with a reduction in long-term mortality, but did not affect the risk of reinfarction and had only a borderline effect on the primary composite end point of death, reinfarction, and class IV heart failure, largely driven by its association with heart failure alone. Although the presence and degree of collaterals was a correlate of improved clinical outcomes, it was not an independent predictor of clinical outcomes, indicating that the major differences in clinical characteristics at baseline associated with different angiographic collateral grades are largely responsible for the subsequent differences in outcomes. Indeed, higher collateral grades were clearly associated with lower risk characteristics at the time of randomization in the subacute phase after infarction. It is likely that higher collateral grade was responsible for some of these lower risk characteristics, for example, improved ejection fraction and less HF, before randomization (ie, during the evolving acute MI), by providing flow to the infarct territory despite persistent occlusion of the infarct artery. They may also be a marker of preinfarct ischemia (stimulating their development) and ischemic preconditioning, either of which could contribute to the apparent protective effect they appear to confer by collaterals. Indeed, there is a host of data that demonstrate that the presence of collaterals during acute MI is associated with a reduction in infarct size and improved left ventricular function.^1–5 However, although collaterals may protect the myocardium against ischemia during an evolving acute MI, the presence of collaterals after the acute phase but before randomization was not an independent predictor of subsequent clinical outcomes, once infarction was completed. This implies that the unadjusted relationship we observed between baseline collaterals and favorable clinical outcome is largely confounded.

It is conceivable that the benefit of late recanalization of an occluded infarct artery may be greatest in patients without collateral flow to the infarct vessel territory. Therefore, it is important to examine a potential interaction between the presence of collaterals at baseline and the impact of PCI (compared with medical therapy alone) on clinical outcomes. In this analysis, the presence of collaterals at baseline angiography did not affect the outcome of the randomized comparison, and there was no interaction between treatment assignment in the trial and angiographic collaterals in terms of any of the end points. Although there was a borderline interaction for HF, suggesting that PCI compared with medical therapy alone may be protective against subsequent development of nonfatal HF in patients without collaterals, it was no longer significant when fatal events were factored in. Thus the potential benefit of PCI in patients without collaterals may be offset by an excess of fatal events, possibly due to the catastrophic consequences of subsequent reocclusion of the recanalized infarct artery in patients without collaterals. Angiographic data from studies such as the DECOPI (DEsobstruction COronaire en Post-Infarctus) trial or TOSCA-2 (Total Occlusion Study of Canada 2) have shown that reocclusion is not uncommon after late recanalization of the infarct artery.^19,20

It is intriguing to think that patients without angiographically visible collaterals to the occluded IRA may derive greater clinical benefit from recanalization of the IRA by PCI than patients who exhibit collaterals to support the infarcted area with an alternative source of blood supply to regions usually supplied by the epicardial vessels. Previous studies have shown that after recanalization of a total chronic occlusion, functional collaterals regress immediately and are even further attenuated in long-term follow-up.²¹ This may expose areas supplied by baseline collaterals to reinfarction in case of acute reocclusion of the recanalized artery, because only a minority of patients will have sufficient and immediately recruitable collaterals in such cases.¹³ Therefore, our observation that the presence of angiographic collaterals at baseline angiography provided no protection against the risk of reinfarction in the present study may be partly explained by the possibility that such collaterals regress and attenuate substantially after recanalization of a total occlusion.

The present study has several strengths and limitations. It is the largest analysis addressing the association between angiographically apparent coronary collaterals and outcome, as well as their relevance for guiding revascularization in stable patients with recent MI. Our analysis was not prespecified. Although angiographic assessment of collaterals has been used widely, no single grading system has achieved universal acceptance. There is evidence that epicardial collaterals are an imprecise surrogate of tissue perfusion in the territory subtended by the occluded coronary and that other measures of tissue perfusion, including contrast echocardiography or hemodynamic assessment of collateral circulation, are better correlated to functional recovery.^22,23 However, these methods are impractical in the context of a large international trial or routine clinical practice. Moreover, our results were consistent, regardless of whether collaterals were assessed in binary fashion or in a graded analysis or whether residual antegrade flow was taken into account in a “total flow” analysis. Finally, collaterals were assessed only at baseline before randomization, and subsequent assessment was not available. Although this is the largest study of its type, the group with no angiographically visible collaterals was modest in size (251 patients).

Conclusion

Angiographic evidence of collaterals supplying an occluded IRA was associated with better global and segmental LV function and a lower prevalence of HF after MI. However, collaterals were not an independent predictor of subsequent clinical outcomes. The presence or absence of collaterals did not influence the overall conclusion from the OAT that late recanalization of an occluded infarct artery does not reduce death, reinfarction, or class IV CHF. Decisions about revascularization in patients with recent MI and occluded IRAs should not be predicated on the presence or grade of angiographic collaterals.

Footnotes