Author + information
- Received December 22, 1998
- Revision received May 4, 1999
- Accepted June 11, 1999
- Published online October 1, 1999.
- William S Weintraub, MD∗,* (, )
- Steven Culler, PhD†,
- Stephen J Boccuzzi, PhD∥,
- John R Cook, PhD∥,
- Andrzej S Kosinski, PhD‡,
- David J Cohen, MD§,
- Joy Burnette, RN∗,
- for the RESTORE Trial Study Group
- ↵*Reprint requests and correspondence: William S. Weintraub, MD, Division of Cardiology, Emory University, 1639 Pierce Drive, WMB 319, Atlanta, GA 30322
This study was conducted to assess the impact of GPIIb/IIIa blockade with tirofiban on costs during the initial hospitalization and at 30 days among patients undergoing high-risk coronary angioplasty.
GPIIb/IIIa blockers are a new class of compounds that have been shown in clinical studies to prevent complications after high-risk angioplasty.
The RESTORE trial was a multinational, blinded placebo-controlled study of 2,197 patients randomized to tirofiban or placebo following coronary angioplasty. This economic assessment was a prospective substudy of the RESTORE trial, and included 1,920 patients enrolled in the U.S. Costs were estimated for the U.S. cohort based on their utilization of healthcare resources and on costs measured directly in 820 U.S. patients at 30 sites.
There was a 36% difference in the rate of the composite event of death, myocardial infarction (MI) and revascularization at two days between tirofiban and placebo (8% vs. 12%, p = 0.002). This difference was attributed to a reduction in nonfatal MI, repeat angioplasty, coronary surgery and stent placement. These clinical benefits followed a similar trend at 30 days, with a 16% reduction in the composite event (p = 0.10). In-hospital cost, including professional and study drug costs, was $12,145 ± 5,882 with placebo versus $12,230 ± 5,527 with tirofiban (p = 0.75). The 30-day cost was $12,402 ± 6,147 with placebo versus $12,446 ± 5,814 with tirofiban (p = 0.87).
Tirofiban has been shown to decrease in-hospital and possibly 30-day events after high-risk angioplasty. The beneficial clinical effects of tirofiban in high-risk patients can be achieved at no increased cost.
The outcome of coronary angioplasty has gradually improved over the last several years. Nonetheless, acute closure remains a significant risk, occurring in some 4% to 13% of patients (1,2). Acute closure is associated with an increased risk of myocardial infarction (MI), emergent coronary surgery and death, resulting in increased resource utilization due to these complications (3). Efforts to decrease acute closure have been aimed partly at mechanical intervention (4), as well as at pharmacologic interventions to prevent thrombosis (5–7). Efforts over the last several years to understand the pathophysiology of the GPIIb/IIIa receptor have revealed the importance of this receptor on the platelet for fibrinogen cross-linking and platelet aggregation (8). Several specific blockers of this receptor have been developed and have been evaluated in clinical studies (9).
In the multicenter Randomized Efficacy Study of Tirofiban for Outcomes and Restenosis (RESTORE) trial, patients were randomized to the specific GPIIb/IIIa receptor blocker tirofiban versus placebo after coronary intervention (7). The RESTORE trial was designed to determine whether tirofiban could prevent untoward complications after coronary intervention at 2, 7 and 30 days postprocedure. As part of the RESTORE trial, a prospective economic analysis of cost was conducted at a subset of sites in the U.S. An examination of the in-hospital and 30-day hospital costs associated with the RESTORE trial is the primary focus of this article.
Between January 1995 and November 1995, 2,197 patients with acute MI (n = 669 nonprimary percutaneous transluminal coronary angioplasty [PTCA] for MI and n = 147 primary PTCA for acute MI) or unstable angina (n = 1,381) undergoing angioplasty were randomized at 111 institutions to tirofiban or placebo. As described in the main report (7), patients <80 years old referred to coronary intervention were eligible if they presented within 72 h of onset of acute MI or with unstable angina. Patients were excluded if they had received thrombolysis within 24 h of the angioplasty, had severe multivessel or left main disease, a contraindication to anticoagulation, severe heart failure or other severe noncardiac problems.
The design and methods for the RESTORE trial have been described previously (7). Briefly, patients received a bolus of 10 μg/kg of tirofiban or placebo over 3 min, after a guide wire was placed across the lesion. Patients then received an infusion of 0.10 μg/kg/min for 36 h. Transcatheter revascularization was performed as clinically indicated. Patients were followed for six months. The primary end point was the composite of death from all causes (acute MI, repeat angioplasty, coronary surgery, stent placement for acute closure or intraaortic balloon counterpulsation to relieve refractory angina) and was evaluated at 30 days. The same end point was also evaluated at two and seven days, as well as at six months. The data were reevaluated in a post hoc analysis, with repeat angioplasty replaced by emergent repeat angioplasty to match similar studies performed with other agents in this class (5,6).
The economic analysis evaluated in this trial included medical costs and resource utilization. In principle, the economic perspective of the study is societal. However, as costs are not accounted for by society, cost models were based on payer data. Hospital charges, collected from UB92 forms, were gathered on 820 patients at 30 U.S. centers. Because of variation in healthcare delivery systems and economics in different countries, the analysis was limited to the U.S.
Hospital costs were estimated for each patient by multiplying total charges by the hospital’s global cost-to-charge ratio from the annual Medicare Cost Report for each hospital (10,11). This methodology, comparing departmental with whole hospital cost-to-charge ratios, was validated in 1,676 patients undergoing angioplasty at Emory University in 1995. If CDis cost from departmental cost-to-charge ratios and CHis cost the whole hospital cost-to-charge ratio, then CH= 0.83 × CD+ 170. The R2was 0.97, and the intercept of 170 was just 2% of the mean whole hospital cost of $8,493. Assuming that CDis a better estimate of cost than CHand that the relationship from Emory between CDand CHholds at other hospitals, then the equation would result in an underestimate of cost for low-cost hospitalizations and an overestimate for high-cost hospitalizations, biasing against a form of therapy that limits costs mostly in higher cost patients by preventing complications.
An estimate of physician costs was derived using a model for physician costs based on a profile of services provided for interventional procedures at Emory University (12). The Emory Clinic’s centralized billing system was used to assess all physician resource use for patients undergoing coronary intervention during any episode of care. The physician cost was derived by converting each CPT code for each service to resource value units (RVUs) (12). The RVUs were then summed and converted to cost using the Medicare conversion factor (12).
The cost of tirofiban was estimated at $700, based on two 50-ml vials of tirofiban injection at a catalogue price of $350 each. Hospital and physician costs for U.S. patients for whom costs were not measured were imputed using a linear regression model with major clinical and outcome variables as independent variables (Table 1). The most important clinical determinants of cost were coronary surgery, additional angioplasty, MI and intracoronary stent placement. A history of congestive heart failure and a MI at presentation were also determinants of cost. Length of stay was included in the model to permit the best possible estimate of cost (R2= 0.57).
Categorical data are presented as counts and percentages. Continuous data are presented as mean ± standard deviation, and as median and interquartile range when the assumption of normality is violated. Categorical data were compared by chi-square or Fisher exact test where appropriate. Continuous data were compared by ttest or by Wilcoxon rank sum test where the assumption of normality is violated. Data on events used for resource utilization in this study will vary somewhat from end point data in the main trial (7)because resource use will consider all events, rather than the somewhat more narrowly defined clinical end points. Data were analyzed with SAS (SAS Institute Inc., Cary, North Carolina) and S-Plus (MathSoft, Seattle, Washington).
The patients in the RESTORE trial (Table 2) were typical of patients undergoing angioplasty in the setting of acute ischemic syndromes. There was no difference noted between the two treatment arms for any baseline clinical variable and no difference between the total group in the U.S. and the patients for whom costs were available. Two- and 30-day clinical outcome data for all U.S. patients are displayed in Table 3. At two days, the composite event rate was 36% lower for tirofiban compared with placebo (8% vs. 12%, p = 0.002). This benefit reflected a reduction in nonfatal MI and repeat angioplasty, as well as trends toward reductions in coronary surgery and stent usage. The data indicated a similar trend in clinical benefit at 30 days, with a 16% reduction in events. When the clinical event data were analyzed for patients included in the cost substudy, the results were similar (Table 4). There was a 46% reduction in events with tirofiban compared with placebo (7.0% vs. 12.9%, p = 0.005). At 30 days, there was a 28% reduction in the composite event rate with tirofiban compared with placebo (12.7% vs. 17.7%, p = 0.049). Direct costs are compared in Table 5. Initial hospital costs, both in those patients for whom costs were measured (n = 820) and in all U.S. patients (with imputed costs where measured costs were not available), were not shown to differ between the two arms. In the patients for whom costs were measured, the mean initial hospital cost was $10,289 with placebo versus $10,551 with tirofiban (p = 0.52). The linear regression model was used to impute costs in all U.S. patients for whom cost data were not collected in RESTORE. There was no significant treatment difference noted in either the initial hospital costs (with the cost of tirofiban included) or the initial professional costs. Thus, the total cost for the initial hospitalization was not shown to differ between the two arms ($12,145 and $12,230 for placebo and tirofiban, respectively). Similarly, at 30 days, the hospital costs and the sum of hospital plus professional costs at $12,402 in the placebo arm and $12,446 in the tirofiban arm were not noted to differ.
The cumulative distribution of initial hospital cost for patients in the cost study is displayed in Figure 1. The cost curves are similar, although there is a suggestion of higher costs in the tirofiban arm for lower cost patients (due to the added cost of tirofiban) and lower cost in the tirofiban arm for higher cost patients (due to fewer complications with tirofiban). This is consistent with tirofiban decreasing cost in higher-cost patients by preventing costly events. The wide dispersion and tail of higher cost patients may be noted. Similarly, the cumulative distribution of hospital cost for all U.S. patients is displayed in Figure 2. While essential overlap exists, it is important to note there is wide variability in cost and some patients at higher cost. Cumulative mean cost over time is displayed in Figure 3, in which the overlap of the two arms may be noted once again.
The nonpeptide GPIIb/IIIa receptor blocker tirofiban has been shown to decrease adverse clinical events after coronary intervention (7)and in the setting of unstable angina (13,14). The clinical results are similar to those noted with the antibody fragment, abciximab (5). In the present study, we have shown that these favorable results can be achieved at essentially no increased cost, due to the decreased cardiovascular events (i.e., a decrease in additional angioplasty, stent placement and coronary surgery). These data may be compared with cost data with abciximab noted in the Evaluation of 7E3 for the Prevention of Ischemic Complications (EPIC) study (15). In EPIC, exclusive of drug cost, the initial cost was $13,577 in the active treatment arm versus $13,434 with placebo, a cost increase of $143 that was related to increased bleeding (14). This compares with an initial hospital and professional cost of $12,230 with tirofiban and $12,145 with placebo in the present study, with the $700 cost of the tirofiban included in the active treatment arm.
Review of GPIIb/IIIa trials
GPIIb/IIIa blockers have established efficacy in decreasing events in the setting of high-risk angioplasty (5–7)and unstable angina (13,14). However, risk stratification to identify the population for whom these agents will prove economically attractive will require additional study (16). While the recently completed Evaluation of PTCA to Improve Long-Term Outcome by C7E3 GPIIb/IIIa Receptor Blockade (EPILOG) (17)and Integrilin to Minimize Platelet Aggregation and Prevent Coronary Thrombosis (IMPACT II) (6)trials suggested that these agents might be useful in wider patient groups, more data are needed to assess the cost-effectiveness of GPIIb/IIIa blockade in lower risk subgroups. Thus, additional studies in low-risk patients may be appropriate.
To date, studies have utilized a bolus or loading infusion followed by a maintenance intravenous infusion. Oral or transdermal GPIIb/IIIa blockers or other antiplatelet agents may extend and improve the utility of intravenous GPIIb/IIIa blockade in the current target populations of unstable angina and high-risk angioplasty. In addition, oral platelet blockade may extend the use of these agents into other patient categories. Given the emerging focus on value in healthcare (cost/outcome), establishing the cost and where appropriate and feasible the cost-effectiveness of these agents will be essential.
Limitations of economic methods in clinical trials
Economic studies within randomized trials have limitations and require careful interpretation. The hospital costs in this study were derived from the UB92 formulation of the hospital bill. The charges on these bills were reduced to costs using whole hospital cost-to-charge ratios. While departmental cost-to-charge ratios may seem more appropriate, there are data that suggest that the whole hospital cost-to-charge ratio is more accurate (18). More importantly, the accuracy of the costs in this study is dependent on the hospital cost reports that are filed annually with the Health Care Financing Agency. Although hospitals have a strong incentive to be accurate in this reporting, it is beyond the ability of this study (or essentially any cost study of this type) to audit the accuracy of cost data.
Professional costs in this trial were based on a model developed in a larger data set at Emory University. While it is easy to criticize these costs as unrepresentative results from one institution, it is probably more meaningful to at least try to account for all professional activity by developing a profile of the distribution of professional activity rather than simply assigning a cost for professional activity to a hospitalization.
It must also be recognized that reimbursement, as well as costs, are changing over time in complex ways. In economic studies, costs are often inflated or deflated to a particular year using the medical inflation rate. However, hospitals have made a great effort to increase efficiency of delivery of health care services, specifically, expensive types of care such as angioplasty. Thus, costs may actually be decreasing while the medical inflation rate is otherwise increasing.
Assessing complete costs has also been a tremendous challenge in medical microeconomic studies. Establishing costs after an initial hospitalization, which may include follow-up hospitalizations at outlying hospitals, is often difficult. Assessing all outpatient direct and indirect costs is difficult and is rarely achieved. Although in cost studies it is generally best to present data from a societal perspective, costs are necessarily accumulated from provider, payer and patient perspectives. Societal costs may be created from these data, recognizing that the data may not always be true societal costs.
In this study, we have shown that the GPIIb/IIIa blocker tirofiban can reduce complications in patients with acute coronary syndromes after angioplasty at no increase in cost, given the limitations of measurement reviewed above. It is difficult for even an effective drug to decrease costs when it decreases the rate of events in only a minority of patients; consequently, cost probably will rise for many patients. However, it will decrease, perhaps dramatically, in patients for whom complications are prevented. Overall, these data are consistent with other agents in the class, and with tirofiban, there was the added benefit that the end point risk reduction was not offset by significant increased bleeding. Finally, the economic analysis in RESTORE suggests the clinical benefit can be achieved at no additional cost in high-risk patients during the initial hospitalization and at 30 days.
We thank Lesley Wood for her careful editorial review.
☆ This work was funded by a grant from Merck & Co., Inc, Whitehouse Station, NJ.
Presented in part at the American College of Cardiology 46th Annual Scientific Session, March 1997.
- Evaluation of 7E3 for the Prevention of Ischemic Complications
- Evaluation of PTCA to Improve Long-Term Outcome by C7ES GPIIb/IIIa Receptor blockade
- Integrilin to Minimize Platelet Aggregation and Prevent Coronary Thrombosis
- myocardial infarction
- percutaneous transluminal coronary angioplasty
- Randomized Efficacy Study of Tirofiban for Outcomes and Restenosis trial
- resource value units
- Received December 22, 1998.
- Revision received May 4, 1999.
- Accepted June 11, 1999.
- American College of Cardiology
- Ellis S.G,
- Roubin G.S,
- King S.B III.,
- Douglas J.S Jr.,
- Weintraub W.S,
- Thomas R.G
- Weintraub W.S,
- Benard J,
- Mauldin P.D,
- Becker E.R,
- King S.B III.
- George B.S,
- Voorhees W.D III.,
- Roubin G.S,
- et al.
- IMPACT investigators,
- Tcheng J.E,
- Harrington R.A,
- Kottke-Marchant K,
- et al.
- RESTORE Investigators
- Weintraub W.S,
- Mauldin P.D,
- Becker E.R,
- Kosinski A.S,
- King S.B III.
- Becker E.R,
- Mauldin P.D,
- Bernadino M.E
- Platelet Receptor Inhibition in Ischemic Syndrome Management in Patients Limited by Unstable Signs and Symptoms (PRISM-PLUS) Study investigators
- Mark D.B,
- Talley J.D,
- Topol E.J,
- et al.
- Weintraub W.S,
- Mahoney E.M
- Ashby J