Author + information
- Received January 23, 2004
- Revision received May 1, 2004
- Accepted May 4, 2004
- Published online November 2, 2004.
- David J. Cohen, MD, MSc*,†,* (, )
- Sabina A. Murphy, MSc*,
- Donald S. Baim, MD‡,
- Tara A. Lavelle, BS*,
- Ronna H. Berezin, MPH*,
- Donald E. Cutlip, MD*,†,
- Kalon K.L. Ho, MD, MSc*,†,
- Richard E. Kuntz, MD, MSc*,‡,
- the SAFER Trial Investigators
- ↵*Reprint requests and correspondence:
Dr. David J. Cohen, Cardiovascular Division, Beth Israel Deaconess Medical Center, 330 Brookline Avenue, Boston, Massachusetts 02215
Objectives The goal of this research was to determine the incremental cost and cost-effectiveness of embolic protection in patients undergoing percutaneous revascularization (PCI) of diseased saphenous vein bypass grafts (SVGs).
Background Distal protection using the GuardWire balloon occlusion device has been shown to reduce major ischemic complications in patients undergoing SVG PCI, but the cost-effectiveness of this approach is unknown.
Methods We prospectively measured medical resource utilization and cost for 801 patients undergoing SVG intervention who were randomized to distal protection using the GuardWire (n = 406) or conventional treatment (n = 395) in the Saphenous Vein Graft Angioplasty Free of Emboli Randomized (SAFER) trial. Long-term survival and cost-effectiveness were projected based on observed 30-day outcomes and a validated survival model for postcoronary artery bypass graft patients.
Results Compared with conventional treatment, distal protection increased initial procedural costs by ∼$1,600 ($6,326 vs. $4,779, p < 0.001). However, by reducing ischemic complications, distal protection reduced mean length of stay by 0.4 days and other hospital costs by nearly $1,000 ($6,846 vs. $7,811, p = 0.018). As a result, overall initial hospital costs were only $582 per patient higher with distal protection. Based on the observed 30-day cost and outcome differences in the trial, the incremental cost-effectiveness ratio for distal protection was $3,718 per year of life saved and remained <$40,000 per year of life saved in 97.3% of bootstrap simulations (95% confidence interval, $0 to $43,079).
Conclusions For patients undergoing PCI of diseased SVGs, distal protection using the GuardWire system is an attractive use of limited health care resources.
Despite major advances in catheter-based therapy and adjunctive pharmacology, percutaneous revascularization of diseased saphenous vein bypass grafts (SVGs) remains a critical challenge for the interventional cardiologist. Vein graft lesions are frequently associated with considerable plaque burden and intracoronary thrombus, which may predispose to complications including distal embolization, the no-reflow phenomenon, and periprocedural myocardial infarction (MI) (1). Moreover, patients with diseased SVGs frequently have additional characteristics including advanced age, more extensive atherosclerosis, and reduced left ventricular function such that they tolerate embolic complications less wellthan a “typical” patient (2,3). Thus, approaches to minimize these complications in patients undergoing vein graft PCI are of paramount importance.
Over the past two decades, a wide range of mechanical and pharmacologic strategies including atherectomy (4), thrombectomy (5,6), intracoronary stenting (7), prolonged infusions of thrombolytic therapy (6), and glycoprotein IIb/IIIa antagonists (8) have been attempted for patients undergoing SVG intervention, with limited success. Recently, however, distal embolic protection using a novel balloon-occlusion device (PercuSurge GuardWire, Medtronic Inc., Minneapolis, Minnesota) has been shown to substantially reduce major complications in this patient population (9).
Although the GuardWire is relatively expensive ($1,500 per device), embolic complications are frequent and costly in patients undergoing SVG percutaneous coronary intervention (PCI) (10). Thus, it is possible that avoidance of such events could offset much of the cost of the device. Moreover, even if the full cost of the device were not recouped, embolic protection could be cost-effective if the benefits of such therapy (e.g., reduced mortality and procedure-related MI) were commensurate with the additional cost (11). We, therefore, conducted a prospective economic study in conjunction with the Saphenous Vein Graft Angioplasty Free of Emboli Randomized (SAFER) trial—a randomized clinical trial to evaluate the safety and effectiveness of distal embolic protection in patients undergoing SVG intervention.
Patient population and treatment protocol
Between January 1999 and August 2000, 801 patients undergoing PCI for a stenotic saphenous vein graft were enrolled in the SAFER trial. Details of the study design have been described previously (9). Patients were eligible if they were undergoing planned PCI to an SVG with a reference diameter between 3 and 6 mm. Patients with ongoing MI, ejection fraction <25%, serum creatinine >2.5 mg/dl, or requiring multivessel PCI were excluded. For the first 142 patients enrolled, lesion length was required to be < one-third of total graft length, but no upper limit on lesion length was imposed for the subsequent 659 patients. The study protocol was approved by the institutional review board at each site, and each patient provided informed consent before enrollment.
Patients were randomized to either conventional PCI (control group, n = 395) or PCI using the PercuSurge GuardWire balloon occlusion device (GuardWire group, n = 406), stratified by clinical site and by whether the operator planned to use a glycoprotein IIb/IIIa receptor antagonist. Conventional PCI was performed over a standard angioplasty guidewire using balloon expandable or self-expanding coronary stents. Patients assigned to the GuardWire group underwent PCI using the GuardWire to occlude the distal vessel during balloon angioplasty and stent deployment, followed by aspiration of atherosclerotic debris from the SVG using the Export aspiration catheter (Medtronic Inc.), before the occlusion balloon was deflated and antegrade flow was restored. Additional details of the GuardWire system and its use have been described previously (9).
Assessment of in-hospital outcomes and clinical follow-up
Case report forms concerning baseline demographic and clinical data, procedural details, and clinical outcomes during the initial hospitalization and 30-day follow-up period were completed by a research coordinator at each site, source-verified by independent data monitors, and submitted to the data coordinating center. All end points (death, MI, repeat revascularization) were reviewed by an independent clinical events committee who were blinded to treatment assignment. Myocardial infarction was defined as elevation of creatinine kinase-MB (CK-MB) >3× the upper limit of normal at any time during the follow-up period. Large MI was defined prospectively as any MI with a peak CKMB >5× the upper limit of normal or any Q-wave MI. Hemorrhagic complications included the need for vascular surgical repair, ultrasound-guided compression, or bleeding requiring transfusion.
Summary of principal clinical outcomes
As reported previously, the primary end point of the SAFER trial was the composite of death, MI, or repeat revascularization of the target vessel at 30 days, which was reduced from 16.5% in the control group to 9.6% in the GuardWire group (42% relative risk reduction, p = 0.004) (9). This difference was driven primarily by a 42% reduction in the incidence of MI with the GuardWire (8.6% vs. 14.2%, p = 0.008). There were also parallel trends toward reduced 30-day mortality (1.0% vs. 2.5%, p = 0.11) and repeat bypass surgery (0.0% vs. 0.5%, p = 0.24) with the GuardWire.
Determination of medical care costs
Medical care costs for the initial hospitalization and for the 30-day follow-up period were assessed using a combination of “bottom-up” and “top-down” methods as previously described (12).
Cardiac catheterization laboratory costs
Detailed resource utilization was recorded for each procedure, and the cost of each item was estimated based on the mean hospital acquisition cost for the item in 2001. Costs of additional disposable equipment, overhead, and depreciation for the cardiac catheterization laboratory, and nonphysician personnel were estimated on the basis of the average cost per procedure at Beth Israel Deaconess Medical Center in 2001 and adjusted for actual procedure duration. The cost of each GuardWire device was set at $1,500 based on its current sales price.
Other hospital costs
All other hospital costs were determined using “top-down” accounting methods based on each hospital's annual Medicare cost report. Itemized bills were obtained for each patient's initial hospitalization and any subsequent cardiovascular hospitalizations during the follow-up period. Hospital costs were determined by multiplying itemized hospital charges by the cost-center specific cost-to-charge ratio obtained from the hospital's Medicare cost report. Previous studies from our group and others have shown this method to correlate well with data from detailed cost accounting systems (13,14). All costs were converted to 2001 dollars based on the medical care component of the Consumer Price Index.
For those admissions with missing billing information (n = 123, 14%), nonprocedural hospital costs were imputed based on a linear regression model developed using the hospital admissions for which complete billing information were available (n = 776). Independent variables for this model included age, length of stay, intensive care unit length of stay, bleeding complications, revascularization procedures, and diagnostic catheterization (model R2= 0.75).
Utilization of selected outpatient services during follow-up (physician office visits, emergency department visits, echocardiograms, and stress tests) was estimated by patient self-report, and costs for these services were calculated based on 2001 Medicare reimbursement rates. Physician's fees for inpatient and outpatient services, major cardiac procedures, and surgical procedures were based on the 2001 Medicare fee schedule.
Discrete data are reported as frequencies, while continuous data are reported as mean ± SD. Cost data are reported as both mean and median values. Discrete variables were compared by Fisher exact test. Normally distributed continuous variables were compared by Student ttest. Cost and other nonnormally distributed data (length of stay, procedure duration) were compared by the Wilcoxon rank-sum test. All statistical analyses and cost-effectiveness analyses were performed according to the intention-to-treat principle.
Because use of the GuardWire was the more costly strategy, we performed a formal cost-effectiveness analysis to compare the clinical benefits with the net cost of distal protection. Because empiric outcomes data were only collected for the 30-day follow-up period, we developed a probabilistic model to project long-term survival beyond the study observation period, contingent on the observed 30-day outcomes. Details of the model are described in the Appendix. The primary end point for the cost-effectiveness analysis was the incremental cost per year of life gained for PCI using the GuardWire compared with conventional PCI. This cost-effectiveness ratio was calculated by dividing the difference in mean 30-day medical care costs for the two treatment groups by the difference in life expectancy, derived from the long-term survival model. We used identical methods to estimate differences in cost, life-expectancy, and incremental cost-effectiveness for several prespecified subgroups according to angiographic, clinical, and treatment-specific factors. Bias-corrected confidence intervals for the cost-effectiveness ratios were estimated by the bootstrap method, using 1,000 resamplings of the study population (15). Although future costs beyond the 30-day trial period were not included in our primary analysis (16,17), additional costs of $2,635 per year (18) were considered in sensitivity analyses. In addition to these long-term analyses, we calculated a “within-trial” cost-effectiveness analysis in which the cost-effectiveness ratio was expressed as cost per death or MI avoided (based on the primary study end points).
Baseline characteristics of the two treatment groups were well-matched (Table 1).The mean age was 69 ± 10 years. Approximately one-third of the population had diabetes mellitus, and more than 90% had multivessel coronary disease. The mean ejection fraction was mildly reduced at 48 ± 12%. The median graft age was 10 years, and ∼40% of lesions contained angiographically evident thrombus.
Procedural resource utilization and cost
Table 2summarizes resource utilization and cost for the index revascularization procedures. Mean procedure duration was increased by 8 min for patients assigned to the GuardWire compared with the control group. Other than the GuardWire itself, there were no other differences in procedural resource utilization between the two treatment groups including the numbers of balloon catheters, stents, and the use of planned or bailout glycoprotein IIb/IIIa inhibitors. Initial procedural costs were ∼$1,600 per patient higher for the GuardWire group compared with the control group ($6,326 ± $1,873 [median $6,012] vs. $4,779 ± $1,771 [median $4,414], p < 0.001)—driven primarily by the cost of the GuardWire itself.
Initial hospital outcomes, resource utilization, and costs.
Table 3summarizes initial hospital outcomes for the two treatment groups. Although there was no significant difference in in-hospital mortality between the GuardWire and control groups (0.7% vs. 1.0%, p = 0.72), patients randomized to the GuardWire had a significantly lower incidence of periprocedural MI during the index hospitalization (8.4% vs. 13.9%, p = 0.01). Although the relative risk reductions were similar for small and large MI (48% and 42%, respectively), approximately two-thirds of the absolute risk reduction for MI was due to prevention of large MI (CK-MB >5× upper limit of normal). In addition, the GuardWire strategy was associated with a modest reduction in the incidence of bleeding complications compared with conventional treatment (5.4% vs. 7.1%, p = 0.38). These lower rates of ischemic and hemorrhagic complications led to a significant reduction in postprocedure length of stay for the GuardWire group compared with the control group (2.7 vs. 3.0 days, p = 0.003).
As a result, hospital room and ancillary costs were ∼$900 per patient lower for the GuardWire group compared with conventional treatment ($4,647 vs. $5,514, p = 0.02). There were no significant differences in either repeat procedure costs or physician costs during the index hospitalization. When combined with the higher initial procedure costs, the net effect of the GuardWire on hospital costs was an increase of ∼$600 per patient ($13,172 ± $5,665 [median $11,639] vs. $12,590 ± $6,252 [median $10,770], p < 0.001).
Follow-up resource utilization and costs
Between hospital discharge and 30-day follow-up, there were trends toward reduced rates of death (0.3% vs. 1.6%, p = 0.06) and nonfatal MI (0.5% vs. 1.6%, p = 0.17) in the GuardWire group (Table 4).All deaths (both in-hospital and during follow-up) were adjudicated as cardiac by the clinical events committee. When these results were combined with the observed in-hospital outcomes, use of the GuardWire was associated with a 42% reduction in the 30-day risk of death or any MI (95% confidence interval, 17% to 65%) and a 44% reduction in the risk of death or large MI (95% confidence interval, 10% to 68%). There were no significant differences in rates of rehospitalization or repeat revascularization procedures during follow-up, however. Follow-up medical care costs for revascularization procedures, hospital services, and outpatient care were virtually identical for the two treatment groups, and cumulative 30-day costs remained $625 per patient higher for the GuardWire group compared with conventional therapy ($14,399 ± $6,731 [median $12,259] vs. $13,774 ± $7,016 [median $11,534], p = 0.006).
To define more precisely those factors that contributed to the net cost of GuardWire treatment, we used multivariable linear regression to determine the impact of specific clinical outcomes, complications, and other factors on initial hospital costs (Table 5).The model identified both procedure-related complications (death, large MI, hemorrhagic complications) and the need for unplanned bypass surgery as the principal determinants of hospital cost with incremental costs ranging from $6,671 (for each major vascular complication) to $46,426 (for in-hospital CABG).
We then estimated the absolute cost savings associated with prevention of specific clinical events by multiplying the independent cost of each event by the difference in event frequency between the GuardWire and conventional treatment groups. Of the $965 per patient savings in nonprocedural costs observed with GuardWire therapy, $310 were attributable to the observed reduction in periprocedural MI, $113 were attributable to reduced bleeding complications, and $83 were attributable to reduced mortality. An additional $237 in cost savings were related to the observed reduction in in-hospital bypass surgery with GuardWire therapy. The remaining $222 in cost savings were not explained by differences in in-hospital events and may reflect differences in the intensity of care or severity of complications between the two treatment groups.
Based on the observed 30-day clinical outcomes and costs, the within-trial cost-effectiveness ratio for GuardWire-based PCI compared with standard PCI was $9,342 per death or MI avoided. On the basis of these results and projections from our long-term survival model, we estimated a mean life expectancy (undiscounted) of 11.38 years for the GuardWire group and 11.16 years for the conventional therapy group—a difference of 0.22 years. After discounting at 3% per year, the life expectancy difference was reduced to 0.17 years (95% confidence interval, 0.04 to 0.29 years). Thus, the lifetime incremental cost-effectiveness ratio for GuardWire-based PCI compared with standard PCI without embolic protection was $3,718 per year of life gained (95% confidence interval, $0 to $43,079). Bootstrap simulation demonstrated that the cost-effectiveness ratio for embolic protection remained <$40,000 per life-year gained in 97.3% of samples (Fig. 1).
These results were relatively stable over a broad range of alternative assumptions regarding long-term outcomes. For example, if we assumed that even large nonfatal MIs had no prognostic significance, the life expectancy gain fell to 0.142 years with an incremental cost-effectiveness ratio of $4,401 per life-year gained. On the other hand, if we assumed that the GuardWire had no effect on 30-day mortality, the life expectancy gain decreased to 0.029 years with an incremental cost-effectiveness ratio of $21,551 per life-year gained. Inclusion of future costs increased the cost-effectiveness ratio to $8,697 per life-year gained, but it remained <$40,000 per life-year gained in 96.5% of bootstrapped samples. Varying the discount rate from 0% to 10% had little effect on the cost-effectiveness ratio (range $2,915 to $5,619 per life-year gained). Finally, varying the relative risk of death or large MI in a deterministic model based on the 30-day SAFER results demonstrated that the cost-effectiveness ratio for embolic protection remained <$40,000 per life-year gained over a broad range of assumptions encompassing the full 95% confidence interval for the treatment effect (Fig. 2).
Stratified analyses according to selected baseline patient characteristics are summarized in Table 6.There were no significant interactions between 30-day outcomes (death or large MI) or costs and treatment assignment for each of the prespecified subgroups (all p > 0.05). While use of the GuardWire was economically dominant for patients with definite angiographic thrombus and for patients with focal lesions (length <10 mm), the cost-effectiveness ratio for embolic protection remained <$20,000 per year of life saved for each of the subgroups examined. Given the reduced sample sizes inherent in subgroup analyses, these cost-effectiveness ratios were less stable than the overall trial results. Nonetheless, the probability that the cost-effectiveness ratio was <$50,000 per life-year gained was >70% for each of the subgroups tested.
Although the safety and effectiveness of embolic protection for patients undergoing SVG PCI have been established (9), the overall cost and cost-effectiveness of such adjunctive therapy were previously unknown. In this prospective economic study, we found that use of the GuardWire embolic protection device increased initial PCI costs by ∼$1,600 per patient compared with standard of care. However, by reducing major ischemic complications (which were both common and costly in this patient population), nearly 60% of this initial cost difference was recouped over the 30-day follow-up period. As a result, overall 30-day medical care costs were increased by <$650 per patient with use of the GuardWire.
Whether this modest cost increase is warranted depends on the extent to which embolic protection improves either long-term survival or quality of life compared with standard of care. Given the relatively short time-frame encompassed by the SAFER trial, we did not believe that there would be sufficient time for quality-of-life differences to emerge. Thus, quality-of-life was not assessed for the trial participants. On the other hand, embolic protection reduced the incidence of 30-day death or MI by 42% (with consistent relative reductions in both death and large MI). When these results were incorporated into a statistical survival model, use of the GuardWire was projected to increase life expectancy by an average of 0.22 years compared with standard care, with a cost-effectiveness ratio of $3,700 per year of life saved. Thus, the cost-effectiveness of embolic protection for patients undergoing SVG PCI compares very favorably with many accepted medical interventions including lipid-lowering therapy for secondary coronary prevention (∼$4,000 per year of life saved) (19), bypass surgery for left main disease (∼$10,000 per year of life saved) (20), and chronic hemodialysis for end-stage renal disease (∼$50,000 per year of life saved) (21). The results of this study, therefore, suggest that distal embolic protection is highly cost-effective for patients undergoing SVG PCI and should not be withheld from such patients on the basis of cost.
Although our cost-effectiveness analysis was based on statistical projections of long-term survival and life expectancy, our model was based on several assumptions that may have biased our results to some extent against the GuardWire. First, although several observational studies have suggested that even small periprocedural MIs are associated with a worse prognosis after PCI (especially among patients with severe underlying coronary artery disease and reduced left ventricular function) (22), we assumed that only larger periprocedural MIs (with CK-MB elevations >5× the upper limit of normal) would adversely affect long-term prognosis after PCI. Second, we assumed that the adverse prognosis associated with periprocedural MI was limited to the first year of follow-up, beyond which survival curves would remain parallel. To the extent that previous studies have generally demonstrated continued divergence of the survival curves beyond the first year of follow-up (23), this assumption of time-limited prognostic significance also represents a conservative assumption that would likely have biased our study against the GuardWire. Finally, we assumed there would be no further differences in cost (among surviving patients) beyond those seen in the first 30 days. However, it is likely that patients with large MIs at the time of PCI would have experienced higher long-term follow-up costs than patients without procedural complications (e.g., for the management of congestive heart failure or arrhythmias). Considering the conservative nature of these assumptions, it is likely that the cost-effectiveness of embolic protection for patients undergoing SVG PCI may be even more favorable than we have estimated.
Role of risk-stratification
With any costly new technology, there is a natural tendency for clinicians as well as policymakers to target the initial use of these therapies to higher risk populations. The value of this approach clearly depends on one's ability to predict reliably the occurrence of complications as well as on the balance of costs and risks associated with the new technology. In the case of embolic protection for SVG PCI, our analysis suggests that efforts to target therapy are unlikely to substantially improve the cost-effectiveness of this technology at the present time. Although one can certainly identify higher and lower risk subgroups within the population of patients undergoing SVG intervention, even the lower-risk populations have a nontrivial probability of major complications and a significant reduction in those complications by use of the GuardWire. As a result, subgroup analyses within the SAFER population demonstrated uniformly favorable cost-effectiveness ratios (although with greater variability due to the reduced sample size).
This study has several important limitations. The primary limitation is the lack of empiric data beyond the 30-day trial period, thus necessitating extensive modeling of long-term survival to develop a meaningful cost-effectiveness analysis. Nonetheless, use of relatively short-term outcomes is common in cardiovascular clinical trials, and many previous cost-effectiveness studies have performed similar extrapolations based on six-month (17,24) or even in-hospital (25,26) data. Moreover, as noted previously, whenever possible we made conservative assumptions so as to bias our analysis againstthe GuardWire. Thus, it is likely that the cost-effectiveness of embolic protection for patients undergoing SVG PCI is even more favorable than we have estimated. Like many device trials, the SAFER trial was conducted at selected sites; whether the clinical and economic outcomes we observed can be generalized to all PCI programs is unknown. Finally, as with any clinical trial, the results of this cost-effectiveness analysis apply only to the study population—patients undergoing SVG intervention—and should not be extrapolated to other populations at high risk of distal embolization such as thrombotic native coronary lesions or patients undergoing primary PCI for acute MI or to other embolic protection devices.
Based on the results of the SAFER trial, distal protection using the GuardWire balloon occlusion system significantly improves the safety of SVG PCI while modestly increasing initial hospital costs and aggregate 30-day costs compared with standard care. Formal economic analysis demonstrates that the cost-effectiveness of the GuardWire in patients undergoing SVG PCI is highly favorable compared with accepted medical interventions and remains reasonable over a wide range of alternative modeling assumptions and patient subgroups. These findings suggest that, for patients undergoing SVG PCI, the GuardWire embolic protection system is an attractive use of scarce societal resources. Further studies are needed to demonstrate the cost-effectiveness of alternative embolic protection devices for SVG PCI, and to extend the results of this study to other PCI populations at high risk of embolic complications.
For the Appendix, please see the November 2, 2004, issue of JACC at www.onlinejacc.org.
Supported by a grant from PercuSurge Corp., a division of Medtronic, Inc.
- Abbreviations and acronyms
- creatinine kinase-MB
- myocardial infarction
- percutaneous coronary intervention
- Saphenous Vein Graft Angioplasty Free of Emboli Randomized trial
- saphenous vein bypass graft
- Received January 23, 2004.
- Revision received May 1, 2004.
- Accepted May 4, 2004.
- American College of Cardiology Foundation
- Hong M.K.,
- Mehran R.,
- Dangas G,
- et al.
- Weintraub W.S.,
- Jones E.L.,
- Morris D.C,
- et al.
- Holmes D.R.,
- Topol E.J.,
- Califf R.M,
- et al.
- Dooris M.,
- Hoffmann M.,
- Glazier S,
- et al.
- Roffi M.,
- Mukherjee D.,
- Chew D.P,
- et al.
- Baim D.S.,
- Wahr D.,
- George B,
- et al.
- Cohen D.J.,
- Krumholz H.M.,
- Sukin C.A,
- et al.
- Efron B.,
- Tibshirani R.J.
- Winkelmayer W.C.,
- Weinstein M.C.,
- Mittleman M.A,
- et al.
- Califf R.M.,
- Abdelmeguid A.E.,
- Kuntz R.E,
- et al.
- Abdelmeguid A.E.,
- Topol E.J.,
- Whitlow P.L.,
- Sapp S.K.,
- Ellis S.G.
- Mark D.,
- Harrington R.,
- Lincoff A,
- et al.
- Weintraub W.S.,
- Thompson T.D.,
- Culler S,
- et al.