Six-month outcomes after percutaneous intervention for lesions in aortocoronary saphenous vein grafts using distal protection devices: Results from the FIRE trial

Article Outline

• Abstract
• Methods
  • Patient population and protocol
  • End points, definitions, and statistical analysis
• Results
  • Patient features and procedural outcomes
  • Six-month outcomes
  • Determinants of MACE after distal protection device–supported PCI
• Discussion
  • Clinical course following SVG intervention with or without distal protection
  • Correlates of late events and implications of procedure-related myonecrosis in distal protection device–supported SVG PCI
  • Study limitations
  • Conclusions and clinical implications
• References
• Copyright

Background

The filter-based FilterWire EX (Boston Scientific, Natick, MA) embolic protection system and the GuardWire (Medtronic, Santa Rosa, CA) balloon occlusion and aspiration device have been previously shown to reduce periprocedural complication rates of percutaneous coronary intervention for saphenous vein graft (SVG) disease and are considered the standard of care in this setting. The late clinical course after treatment with these devices has not been reported.

Methods

In the FIRE trial, 651 patients undergoing SVG intervention were randomized to either the FilterWire EX or GuardWire. Six-month rates of the primary end point (composite major adverse cardiac events [MACE]) and its components (death, myocardial infarction [MI], or target vessel revascularization) were studied.

Results

MACE at 30 days occurred in 9.9% of patients randomized to the FilterWire EX compared with 11.6% with the GuardWire, P = .53. By 6 months, MACE had increased to 19.3% and 21.9% in FilterWire EX and GuardWire groups, respectively, (relative risk 0.88, 95% CI 0.65-1.19; P = .44). All-cause 6-month mortality in the entire population was 3.5% (3.0% with FilterWire EX vs 4.1% with GuardWire, P = .53, with all deaths occurring after hospital discharge). MI occurred in 12.0% of patients at 6 months (12.1% vs 11.9% with the FilterWire EX and GuardWire, respectively, P = .99), and target vessel revascularization was required in 9.1% (8.2% vs 10.0%, respectively, P = .42).

Conclusions

SVG intervention with the FilterWire EX and GuardWire distal protection devices resulted in similar outcomes at 6 months, although the clinical course after hospital discharge was not benign, with significant rates of death, MI, and repeat intervention.

By reducing the risk of procedure-related atheroembolism, the use of a distal protection device has become standard of care during percutaneous coronary intervention (PCI) of diseased aortocoronary saphenous vein bypass grafts (SVGs).¹, ² However, although the routine use of distal protection devices has been found to enhance the short-term safety profile of the procedure,¹, ² whether long-term patient outcomes after SVG angioplasty are also improved has not been reported. In this regard, several studies before the introduction of distal protection devices reported a poor late prognosis after SVG angioplasty, with high rates of death, myocardial infarction (MI), and target vessel revascularization (TVR).³, ⁴, ⁵, ⁶, ⁷ Moreover, whereas prior studies found that periprocedural MI occurring after SVG intervention without distal protection correlates with late mortality,⁸, ⁹ the prognostic implication of procedure-related myonecrosis occurring despite the use distal protection has also not been studied.

The pivotal FIRE trial established the noninferiority of the filter-based FilterWire EX (Boston Scientific, Natick, MA) compared with the balloon occlusion and aspiration GuardWire (Medtronic, Santa Rosa, CA) in terms of preventing 30-day major adverse cardiac events (MACE) after PCI in diseased SVGs.² In the present analysis from FIRE, we (1) examined the 6-month outcomes from this large population of patients undergoing SVG intervention with routine use of distal protection, and (2) sought to determine the clinical and angiographic correlates of late adverse events after distal protection device–supported PCI for SVG disease.

Back to Article Outline

Methods 

Patient population and protocol 

Details of the FIRE trial and the devices used have been described previously.² Briefly, 651 patients undergoing PCI with planned stenting of one or more de novo SVG lesions were randomized to treatment with either the FilterWire EX (n = 332) or the GuardWire (n = 319). Patients were excluded for aorto-ostial lesions, SVG age <6 months, or distal vessel anatomy unsuitable for distal protection device placement. Other exclusion criteria included recent MI or stroke, serum creatinine ≥2.5 mg/dL, left ventricular ejection fraction <25%, or planned use of a thrombectomy device. Protocol-mandated antithrombotic therapy included preprocedural aspirin 325 mg and thienopyridine (either clopidogrel 300 mg or ticlopidine 500 mg). After deployment of the distal protection device, stenting was performed using standard techniques. Use of glycoprotein IIb/IIIa inhibitors and prestent and poststent balloon inflations were left to the discretion of the operator. Quantitative coronary angiography was performed at an independent core angiographic laboratory, using previously validated methodology. Clinical follow-up after discharge was performed at 30 days and 6 months and was available for 624 patients (96%) who comprise the present study population.

End points, definitions, and statistical analysis 

The primary end point was a composite of MACE defined as death, MI, or TVR.² MI was defined as any postprocedural elevation of creatine kinase (CK)–MB isoenzyme >3 times normal (CK-MB ratio >3).² Categorical variables were compared using the χ² test. Continuous variables are presented as means ± SD and were compared with 1-way analysis of variance or unpaired t test. Clinical outcomes are presented as Kaplan-Meier survival estimates and were compared with the log-rank test. Cox proportional hazard regression analysis was performed to identify independent predictors of late clinical events, with all the baseline variables in Table I entered into the model. Postprocedural inhospital MACE complications were also added into the models for postdischarge adverse events, but not for cumulative events. The SVG degeneration score was calculated as the axial length of luminal irregularities or ectasia expressed as a percent of the entire graft length. For all analyses, a 2-sided P < .05 was considered statistically significant.

Table I. Selected baseline features of patients with 6-month follow-up

	N = 624
Clinical features
Age, y	69.5 ± 10.0
Men, %	79.3
Canadian Heart Association angina class, %
I	7.0
II	15.6
III	29.1
IV	45.4
Smoking within 1 y, %	15.4
Hypertension, %	81.5
Hyperlipidemia, %	89.4
Diabetes, %	39.1
History of cerebrovascular accident, %	13.6
Peripheral vascular disease, %	29.6
History of renal failure, %	12.8
Previous MI, %	65.4
Left ventricular ejection fraction, %	48.4 ± 12.2
SVG age, y	10.7 ± 5.4
Aspirin used during the week preceding PCI, %	80.5
Clopidogrel used ≥3 d preceding PCI, %	48.7
Angiographic features
SVG distribution, %
Left anterior descending artery	18.9
Left circumflex artery	40.0
Right coronary artery	40.8
Proximal or ostial lesion location, %	48.6
Lesion length, mm	13.4 ± 8.6
Reference vessel diameter, mm	3.36 ± 0.68
Minimal luminal diameter (preprocedural), mm	1.13 ± 0.48
Diameter stenosis (preprocedural), %	66.2 ± 13.4
TIMI flow (preprocedure), %
0-1	2.55
2	15.6
3	81.9
SVG degeneration score,⁎ %	26. 9 ± 23.8
Procedural features and results
Thienopyridine pretreatment used, %	48.7
Glycoprotein IIb/IIIa antagonist used, %	53.5
Stents implanted, %	98.3
Number of stents implanted (per patient)	1.5 ± 0.90
In-lesion minimal luminal diameter (postprocedural), mm	2.91 ± 0.70
In-lesion diameter stenosis (postprocedural), %	12.9 ± 10.1
In-stent minimal luminal diameter, mm	3.22 ± 0.62
In-stent diameter stenosis, %	4.0 ± 12.8
TIMI flow (postprocedure), %
0-1	0.2
2	3.0
3	96.8

Back to Article Outline

Results 

Patient features and procedural outcomes 

Baseline clinical, angiographic, and procedural characteristics of the entire study population are shown in Table I. The baseline features of patients enrolled in the 2 randomization arms have been previously reported.² Other than small differences in the number of stents used and postprocedural in-lesion diameter stenosis (1.4 ± 0.9 vs 1.6 ± 0.9, and 12.2% ± 9.5% vs 13.7% ± 10.7%, respectively, P = .04 for both comparisons), the baseline characteristics of study patients treated with the FilterWire EX versus GuardWire devices were similar.

Six-month outcomes 

Six-month follow-up data were available for 317 and 307 patients in the FilterWire and GuardWire groups, respectively. Adverse cardiac events rates in the entire study population and among patients stratified by the type of distal protection device used are shown in Table II. Pooled mortality rates were 0% inhospital, 0.9% between discharge to 30 days, and 3.7% cumulative to 6 months. The inhospital, postdischarge, and cumulative 6-month rates of death, MI, TVR, and the composite MACE end points were similar in patients treated with the FilterWire EX and GuardWire devices (Figure 1 and Table II).

Table II. MACE rates at 6-month follow-up

	All patients	FilterWire EX (n = 317)	GuardWire (n = 307)	Relative risk (95% CI)	P
Inhospital event rates, %
Death	0.0	0.0	0.0	N/A	N/A
MI	7.5	7.3	7.8	0.93 (0.54-1.59)	.88
Q-wave MI	0.6	0.6	0.6	0.96 (0.14-6.80)	.99
Non–Q-wave MI	6.9	6.6	7.2	0.92 (0.52-1.62)	.88
TVR	0.8	0.6	0.9	0.64 (0.11-3.82)	.68
Stroke	0.5	0.3	0.6	0.48 (0.04-5.29)	.62
Composite end point	7.7	7.6	7.8	0.96 (0.57-1.64)	.99
Out-of-hospital event rates, %
Death	2.8	2.4	3.3	0.68 (0.26-1.78)	.43
MI	3.2	3.8	2.6	1.46 (0.60-3.57)	.41
Q-wave MI	0.0	0.0	0.0	N/A	N/A
Non–Q-wave MI	3.2	3.8	2.6	1.46 (0.60-3.57)	.41
TVR	8.7	8.0	9.4	0.83 (0.48-1.41)	.49
Stroke	0.5	0.3	0.7	0.48 (0.04-5.35)	.55
Composite end point	12.5	11.1	13.9	0.78 (0.50-1.22)	.27
Cumulative event rates, %
Death	3.7	3.3	4.2	0.74 (0.33-1.70)	.43
MI	12.2	12.0	12.1	1.01 (0.65-1.58)	.95
Q-wave MI	0.8	0.9	0.6	1.44 (0.24-8.64)	.68
Non–Q-wave MI	11.7	11.6	11.7	0.99 (0.63-1.55)	.96
TVR	9.5	8.6	10.4	0.81 (0.48-1.35)	.41
Stroke	1.1	0.9	1.3	0.72 (0.16-3.23)	.77
Composite end point	20.9	19.7	22.2	0.87 (0.62-1.23)	.43

• 
  • View Large Image
  • Download to PowerPoint

• Figure 1. 

  Freedom from 6-month MACE in patients treated either with the FilterWire EX (solid line) or the GuardWire (broken line) devices.

Determinants of MACE after distal protection device–supported PCI 

Multivariate correlates of adverse outcomes occurring after hospital discharge to 6-month follow-up and cumulative to 6 months are listed in Table III. Mortality was independently predicted by proximal location of the target lesion, TIMI 0/1 flow in the treated graft before PCI, and reduced baseline left ventricular function. Inhospital MI was predicted by both lesion length and SVG degeneration score, although no predictive factors were identified for the occurrence of out-of-hospital MI. Device use, including the number and length of stents deployed and the type of distal protection used, was not an independent correlate of 6-month MACE.

Table III. Multivariate correlates of adverse outcomes

	Hazard ratio	95% CI	P
Inhospital events
Death⁎	–	–	–
MI
SVG degeneration score†	1.02	1.01-1.03	.0008
Lesion length	1.03	1.01-1.06	.01
TVR
Baseline reference vessel diameter	23.07	2.55-208.5	.005
SVG degeneration score†	1.06	1.01-1.10	.009
Composite MACE end point
SVG degeneration score†	1.02	1.01-1.03	.006
Lesion length†	1.03	1.01-1.06	.02
Baseline reference vessel diameter	1.56	1.02-2.39	.04
Out-of-hospital 6-m events‡
Death
Proximal lesion location	4.34	1.51-12.50	.007
Preprocedure TIMI flow 0/1	11.09	1.29-95.58	.03
Baseline left ventricular ejection fraction†	0.95	0.91-0.99	.03
Postprocedural CK-MB ratio >3	1.98	0.41-9.65	.40
MI§
Postprocedural CK-MB ratio >3	1.30	0.29-5.81	.73
TVR
Postprocedural CK-MB ratio >3	2.58	1.15-5.77	.02
Age†	0.97	0.94-0.99	.03
Composite MACE end point
Postprocedural CK-MB ratio >3	1.99	0.97-4.13	.06
Cumulative 6-m events
Death‖
Proximal lesion location	4.34	1.51-12.50	.007
Preprocedure TIMI flow 0/1	11.09	1.29-95.58	.03
Baseline left ventricular ejection fraction†	0.95	0.91-0.99	.03
MI
SVG degeneration score†	1.01	1.01-1.02	.007
Lesion length	1.03	1.01-1.05	.03
TVR
Age†	0.01	0.94-0.99	.01
Diabetes	0.50	0.26-0.98	.04
SVG degeneration score†	1.01	1.00-1.02	.04
Composite MACE end point
Lesion length†	1.03	1.01-1.04	.005
SVG degeneration score†	1.01	1.00-1.02	.02

Postprocedural MI was not a statistically significant predictor of late mortality either by univariate or multivariate analysis (Figure 2 and Table III). Procedure-related MI did correlate with out-of-hospital TVR, however, along with younger age.

• 
  • View Large Image
  • Download to PowerPoint

• Figure 2. 

  Six-month out-of-hospital MACE rates in patients stratified by postprocedural CK-MB levels. Time point 0 signifies hospital discharge. P values are for comparisons between patients with (dashed line) versus without (solid and dotted lines) postintervention CK-MB ratio >3.

Back to Article Outline

Discussion 

The main findings of the present study are that (1) the noninferiority of the filter-based FilterWire EX compared with the GuardWire balloon occlusion and aspiration distal protection device previously demonstrated at 30 days was maintained at 6-month follow-up; (2) despite reductions in periprocedural complications with use of distal protection devices during PCI of diseased SVG, the absolute out-of-hospital rates of death, MI, and TVR were significant, and (3) periprocedural MI occurring after SVG intervention with distal protection use was not strongly related to mortality or MACE at 6 months.

Clinical course following SVG intervention with or without distal protection 

Due in part to the increased morbidity and mortality of repeat bypass surgery, PCI had become the preferred therapeutic strategy for many patients with SVG disease even before the introduction of distal protection devices.¹⁰, ¹¹ Although long-term freedom from reintervention after SVG angioplasty has been gradually improving as a result of the widespread utilization of stents,⁴ recent studies in patients not treated with distal protection have still reported high 1- and 3-year rates of mortality (6% and 9%, respectively) and MACE (30% and 43%, respectively).⁴, ⁵ By reducing the occurrence of procedure-related distal embolization and periprocedural MI, distal microcirculatory protection with either the FilterWire EX or GuardWire has become established as the standard of care during SVG intervention.¹, ² However, long-term outcomes after SVG intervention performed with distal protection devices have not been reported previously, so whether the beneficial impact of these devices on procedural safety translates into a more benign postdischarge clinical course is unknown.

In the present study in which all patients were treated with distal protection and most (98%) with bare metal stent implantation, the incidence of out-of-hospital MACE at 6 months was significant: the rates of the primary composite end point, which were similar in the FilterWire EX and GuardWire groups, approximately tripled between discharge and 6-month follow-up (from 7.7% to 20.6%). These findings demonstrate that although distal protection devices enhance procedural success rates and short-term outcomes in catheter-based therapy of SVG disease, these patients remain at high risk for future adverse events after discharge. Of note, in a registry study of patients with SVG disease managed by PCI without the use of distal protection devices, repeat TVR procedures accounted for the majority of MACE during the first 6 months after intervention, whereas the rates of MI and death began to rise only after this period.⁵ TVR was also the most frequent adverse event during the first 6 months after the procedure in the randomized FIRE trial, although the out-of-hospital MI rate (∼5%) was higher than reported previously in nonrandomized studies,⁴, ⁵ possibly as a result of increased surveillance for MI in the present study or other patient-related or methodological differences between the reports.

Correlates of late events and implications of procedure-related myonecrosis in distal protection device–supported SVG PCI 

In the present study, all patient deaths occurred after hospital discharge. Multivariate analysis identified reduced baseline left ventricular ejection fraction and 2 angiographic variables—proximal lesion location and occlusion of the target graft before PCI—as independent predictors of out-of-hospital (and therefore cumulative) 6-month mortality after hospital discharge. Cumulative 6-month MI rates were predicted by increasing SVG degeneration score and lesion length. These results are consistent with previous experience with SVG PCI without distal protection, emphasizing the importance of angiographic lesion characteristics in determining SVG-related myonecrosis, whether or not distal protection is used.⁴ Also consistent with previous findings, there was a trend toward less TVR with increasing age and the presence of diabetes, possibly reflecting referral bias of elderly patients and those with existing comorbidities for repeat high-risk interventions.⁴

In the present study, the only late adverse event that correlated strongly with procedure-related MI was the need for repeat out-of-hospital TVR. Although the mechanism underlying this association is uncertain, periprocedural cardiac enzyme elevation, despite the use of a distal protection device, may be a marker for patients with either increased or friable atheromatous burden at and beyond the intervention site, which may also predict accelerated SVG degeneration and subsequent TVR. Conversely, the absence of an independent association between periprocedural myonecrosis and late mortality stands in contrast to our previous findings in patients with SVG disease treated by PCI without distal protection,⁸ in which the 6-month and 1-year mortality rates were approximately 4-fold greater in patients with a procedure-related MI compared with those with an uncomplicated procedure. Moreover, given the efficacy of distal protection in reducing periprocedural complications, the 3.5% postdischarge mortality rate at 6 months was surprisingly high. Although large procedure-related MIs can cause irreversible left ventricular damage, thus impacting survival after PCI in either native or SVG disease,¹², ¹³ lesser cardiac enzyme elevations (representing most procedure-related MIs) are of reduced prognostic significance after PCI,¹², ¹³ but may serve as a marker for more extensive atherosclerosis.¹⁴, ¹⁵ Thus, the prevention of procedure-related myonecrosis may only partially improve the otherwise poor long-term prognosis of patients with degenerated SVGs requiring revascularization.

Study limitations 

The present study represents the largest population of patients after SVG intervention with distal protection in whom systematic follow-up was performed within the construct of a closely monitored and adjudicated randomized clinical trial. Nonetheless, several limitations need to be acknowledged. First, the number of procedure-related MI events in the FIRE trial was relatively small, and the present analysis was not designed nor powered to detect slight differences in MACE with respect to postprocedural CK-MB elevation. Furthermore, as seen in Figure 2, there is a suggestion that late mortality beyond 6 months may be beginning to rise in patients with periprocedural MI. Thus, larger studies with longer-term follow-up are required before an association between postprocedure adverse events after SVG intervention with distal protection devices and survival can be excluded. Second, given the lack of a control arm, the results of the present study should not be taken as evidence that distal protection devices do not improve long-term outcomes after SVG intervention. Indeed, late mortality may have been even greater had SVG intervention been undertaken without distal protection. Long-term data from the only randomized placebo-controlled distal protection device study (the SAFER trial¹), if available, might provide important insight into this issue. Third, revascularization events in nontarget vessels were not captured in FIRE; thus, the true rate of any repeat intervention was likely underestimated in the present study. Finally, comprehensive data on medications prescribed at discharge, which potentially might impact late prognosis (eg, statins and β-blockers), were unavailable, although most likely were equally represented in both groups given the randomization process.

Conclusions and clinical implications 

The present study demonstrates that the performance of PCI in diseased SVGs with distal protection using either the filter-based FilterWire EX or the GuardWire balloon occlusion and aspiration system results in similar outcomes both in the periprocedural period and for 6 months thereafter. Distal protection devices have been shown to enhance procedural success rates, reduce the occurrence of no reflow, and prevent large as well as small periprocedural infarctions¹; the former of which is clearly prognostically relevant.¹², ¹³ These results thus support the general recommendation for routine use of distal protection devices during SVG intervention when possible. However, it should be recognized that the long-term course after intervention in diseased SVGs is not benign even when distal protection is used. Periprocedural MACE still occurs in approximately 10% of patients, and the rates of death, MI, and repeat revascularization procedures are relatively high compared with PCI in native coronary arteries, due not only to restenosis at the target site but also to disease progression in sites remote from that of the index intervention.⁵ These observations underscore the need for intensive postdischarge surveillance and secondary prevention measures in this population, as well as further research into novel pharmacological and interventional strategies to enhance event-free survival in patients with SVG disease managed by PCI.

Back to Article Outline

References 

1. Baim DS, Wahr D, George B, et al. Randomized trial of a distal embolic protection device during percutaneous intervention of saphenous vein aorto-coronary bypass grafts. Circulation. 2002;105:1285–1290
   • View In Article
2. Stone GW, Rogers C, Hermiller J, et al. Randomized comparison of distal protection with a filter-based catheter and a balloon occlusion and aspiration system during percutaneous intervention of diseased saphenous vein aorto-coronary bypass grafts. Circulation. 2003;108:548–553
   • View In Article
   • CrossRef
3. Savage MP, Douglas JS, Fischman DL, et al. Stent placement compared with balloon angioplasty for obstructed coronary bypass grafts. Saphenous Vein De Novo Trial Investigators. N Engl J Med. 1997;337:740–747
   • View In Article
   • MEDLINE
   • CrossRef
4. Hong MK, Mehran R, Dangas G, et al. Are we makinh progress with percutaneous saphenous vein graft treatment? A comparison of 1990 to 1994 and 1995 results. J Am Coll Cardiol. 2001;38:150–154
   • View In Article
   • Abstract
   • Full Text
   • Full-Text PDF (144 KB)
   • CrossRef
5. Keeley EC, Velez CA, O'Neill WW, et al. Long-term clinical outcome and predictors of major adverse cardiac events after percutaneous interventions on saphenous vein grafts. J Am Coll Cardiol. 2001;38:659–665
   • View In Article
   • Abstract
   • Full Text
   • Full-Text PDF (195 KB)
   • CrossRef
6. Nishida T, Colombo A, Briguori C, et al. Contemporary percutaneous treatment of saphenous vein graft stenosis: immediate and late outcomes. J Invasive Cardiol. 2000;12:505–512
   • View In Article
   • MEDLINE
7. Hanekamp CE, Koolen JJ, Den Heijer P, et al. Randomized study to compare balloon angioplasty and elective stent implantation in venous bypass grafts: the Venestent study. Catheter Cardiovasc Interv. 2003;60:452–457
   • View In Article
   • MEDLINE
   • CrossRef
8. Hong MK, Mehran R, Dangas G, et al. Creatine kinase–MB enzyme elevation following successful saphenous vein graft intervention is associated with late mortality. Circulation. 1999;100:2400–2405
   • View In Article
9. Wolfram R, Leborgne L, Cheneau E, et al. Comparison of creatine kinase elevation and outcome of comparison of percutaneous coronary intervention for saphenous vein graft in-stent restenosis versus de novo stenosis. Am J Cardiol. 2003;92:980–983
   • View In Article
   • Abstract
   • Full Text
   • Full-Text PDF (112 KB)
   • CrossRef
10. Stephan WJ, O'Keefe JH, Piehler JM, et al. Coronary angioplasty versus repeat coronary artery bypass grafting for patients with previous bypass surgery. J Am Coll Cardiol. 1996;28:1140–1146
    • View In Article
    • Abstract
    • Full-Text PDF (689 KB)
    • CrossRef
11. Morrison DA, Sethi G, Sacks J, et al. Percutaneous coronary intervention versus repeat bypass surgery for patients with medically refractory myocardial ischemia: AWESOME randomized trial and registry experience with post-CABG patients. J Am Coll Cardiol. 2002;40:1951–1954
    • View In Article
    • Abstract
    • Full Text
    • Full-Text PDF (73 KB)
    • CrossRef
12. Stone GW, Mehran R, Dangas G, et al. Differential impact on survival of electrocardiographic Q-wave versus enzymatic myocardial infarction after percutaneous intervention: a device-specific analysis of 7147 patients. Circulation. 2001;104:642–647
    • View In Article
    • CrossRef
13. Ellis SG, Chew D, Chan A, et al. Death following creatine kinase–MB elevation after coronary intervention: identification of an early risk period: importance of creatine kinase–MB level, completeness of revascularization, ventricular function, and probable benefit of statin therapy. Circulation. 2002;106:1205–1210
    • View In Article
    • CrossRef
14. Mehran R, Dangas G, Mintz GS, et al. Atherosclerotic plaque burden and CK-MB enzyme elevation after coronary interventions: intravascular ultrasound study of 2256 patients. Circulation. 2000;101:604–610
    • View In Article
15. Nallamothu BK, Bates ER. Periprocedural myocardial infarction and mortality: causality versus association. J Am Coll Cardiol. 2003;42:1412–1414
    • View In Article
    • Full Text
    • Full-Text PDF (70 KB)
    • CrossRef