economic evaluation of percutaneous coronary intervention in

149
Economic Evaluation of Percutaneous Coronary Intervention in Stable Coronary Artery Disease: Studies in Utilities and Decision Modeling by Harindra Wijeysundera A thesis submitted in conformity with the requirements for the degree of Doctor of Philosophy, Clinical Epidemiology and Health Care Research, Department of Health Policy, Evaluation and Management University of Toronto © Copyright by Harindra Wijeysundera, 2011

Upload: doanxuyen

Post on 02-Jan-2017

217 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Economic Evaluation of Percutaneous Coronary Intervention in

Economic Evaluation of Percutaneous Coronary Intervention in Stable Coronary Artery Disease: Studies in

Utilities and Decision Modeling

by

Harindra Wijeysundera

A thesis submitted in conformity with the requirements for the degree of Doctor of Philosophy, Clinical Epidemiology and Health Care

Research,

Department of Health Policy, Evaluation and Management

University of Toronto

© Copyright by Harindra Wijeysundera, 2011

Page 2: Economic Evaluation of Percutaneous Coronary Intervention in

ii

Economic Evaluation of Percutaneous Coronary

Intervention in Stable Coronary Artery Disease: Studies in

Utilities and Decision Modeling

Harindra Wijeysundera

Doctor of Philosophy

Health Policy, Evaluation and Management

University of Toronto

2011

Abstract

The initial treatment options for patients with stable coronary artery disease include optimal

medical therapy alone, or coronary revascularization with optimal medical therapy. The most

common revascularization modality is percutaneous coronary intervention (PCI) with either bare

metal stents (BMS) or drug-eluting stents (DES). PCI is believed to reduce recurrent angina and

thereby decrease the need for additional procedures compared to optimal medical therapy alone.

It remains unclear if these benefits are sufficient to offset the increased costs and small increase

in adverse events associated with PCI.

The objectives of this thesis were to determine the degree of angina relief afforded by PCI and

develop a tool to provide contemporary estimates of the impact of angina on quality of life. In

addition, we sought to develop a comprehensive state-transition model, calibrated to real world

costs and outcomes to compare the cost-effectiveness of initial medical therapy versus PCI with

either BMS or DES in patients with stable coronary artery disease.

Page 3: Economic Evaluation of Percutaneous Coronary Intervention in

iii

We performed a systematic search and meta-analysis of the published literature. Although PCI

was associated with an overall benefit on angina relief (odds ratio [OR] 1.69; 95% Confidence

Interval [CI] 1.24-2.30), this benefit was largely attenuated in contemporary studies (OR 1.13;

95% CI 0.76-1.68). Our meta-regression analysis suggests that this observation was related to

greater use of evidence-based medications in more recent trials.

Using simple linear regression, we were able to create a mapping tool that could accurately

estimate utility weights from data on the Seattle Angina Question, the most common descriptive

quality of life instrument used in the cardiovascular literature.

In our economic evaluation, we found that an initial strategy of PCI with a BMS was cost-

effective compared to medical therapy, with an incremental cost-effectiveness ratio (ICER) of

$13,271 per quality adjusted life year gained. In contrast, DES had a greater cost and lower

survival than BMS and was therefore a dominated strategy.

Page 4: Economic Evaluation of Percutaneous Coronary Intervention in

iv

Acknowledgments

I acknowledge support from the University of Toronto, Department of Medicine Clinician

Scientist Training Program and a research fellowship award from the Canadian Institute of

Health Research (CIHR). The clinical registry data used in this publication are from the Alberta

Provincial Project for Outcome Assessment in Coronary Heart Disease (APPROACH) database

and the Cardiac Care Network of Ontario and its member hospitals. The Cardiac Care Network

of Ontario serves as an advisory body to the Ministry of Health and Long Term Care

(MOHLTC) of Ontario and is dedicated to improving the quality, efficiency, access and equity of

adult cardiovascular services in Ontario, Canada. The Cardiac Care Network of Ontario is

funded by the MOHLTC. This study was supported by the Institute for Clinical Evaluative

Sciences (ICES), and the Toronto Health Economics and Technology Assessment (THETA)

collaborative, which are funded by an annual grant from the MOHLTC. The opinions, results

and conclusions reported in this paper are those of the authors and are independent from the

funding sources. No endorsement by ICES or the Ontario MOHLTC is intended or should be

inferred.

Finally, I thank my wife, Sophia and children Emiko, Sunil and Amali, for their support and

understanding, without which this work would not be possible.

Page 5: Economic Evaluation of Percutaneous Coronary Intervention in

v

Table of Contents

Table of Contents

Acknowledgments .......................................................................................................................... iv

Table of Contents ............................................................................................................................ v

List of Tables ............................................................................................................................... viii

List of Figures ................................................................................................................................ ix

List of Appendices .......................................................................................................................... x

1 Introduction ................................................................................................................................ 1

1.1 Contemporary Treatment Options: PCI or Medical Therapy ............................................. 1

1.2 Evaluation of Angina Relief Associated with PCI ............................................................. 2

1.3 Impact of Angina on Quality of Life .................................................................................. 5

1.4 Evaluating trade-offs: a decision analytic framework ........................................................ 7

1.5 Literature to date ................................................................................................................. 8

1.6 Gaps in Knowledge ........................................................................................................... 10

2 Effects of PCI versus Medical Therapy on Angina Relief: A Meta-analysis .......................... 13

2.1 Introduction ....................................................................................................................... 16

2.2 Methods ............................................................................................................................. 17

2.2.1 Data Sources and Searches ................................................................................... 17

2.2.2 Study Selection ..................................................................................................... 17

2.2.3 Data Extraction ..................................................................................................... 18

2.2.4 Data Synthesis and Analysis ................................................................................. 18

2.2.5 Role of the Funding Source .................................................................................. 20

2.3 Results ............................................................................................................................... 21

2.3.1 Study Selection ..................................................................................................... 21

Page 6: Economic Evaluation of Percutaneous Coronary Intervention in

vi

2.3.2 Study Design ......................................................................................................... 21

2.3.3 Freedom from Angina ........................................................................................... 22

2.3.4 Subgroup Analysis on Post Infarction Status ........................................................ 22

2.3.5 Subgroup Analysis on Coronary Stent Use .......................................................... 23

2.3.6 Freedom from Angina According to Enrollment Period ...................................... 23

2.3.7 Meta-Regression of Freedom from Angina and Medical Therapy ....................... 23

2.4 Discussion ......................................................................................................................... 24

3 Predicting EQ-5D Utility Scores from the Seattle Angina Questionnaire in Coronary

Artery Disease: A Mapping Algorithm using a Bayesian Framework .................................... 39

3.1 Introduction ....................................................................................................................... 43

3.2 Methods ............................................................................................................................. 44

3.2.1 Data Source: .......................................................................................................... 44

3.2.2 Measures: .............................................................................................................. 44

3.2.3 Analysis: ............................................................................................................... 45

3.3 Results ............................................................................................................................... 49

3.3.1 Study Cohort ......................................................................................................... 49

3.3.2 Model Estimation .................................................................................................. 50

3.3.3 Model Reliability and Validation .......................................................................... 51

3.3.4 Secondary Analyses .............................................................................................. 52

3.4 Discussion ......................................................................................................................... 52

4 Medical Therapy vs. PCI in Stable Coronary Artery Disease: A Decision Analysis and

Economic Evaluation ............................................................................................................... 63

4.1 Introduction ....................................................................................................................... 66

4.2 METHODS ....................................................................................................................... 67

4.2.1 Research Ethics Board Approval: ......................................................................... 67

4.2.2 Study Design and Outcomes: ................................................................................ 67

4.2.3 Economic Assumptions: ....................................................................................... 67

Page 7: Economic Evaluation of Percutaneous Coronary Intervention in

vii

4.2.4 Base Case: ............................................................................................................. 67

4.2.5 Data Source: .......................................................................................................... 68

4.2.6 Treatment Strategies and Model Structure: .......................................................... 69

4.2.7 Probabilities and Hazard Ratios: ........................................................................... 71

4.2.8 Utilities: ................................................................................................................. 73

4.2.9 Costs: ..................................................................................................................... 74

4.2.10 Calibration: ........................................................................................................... 75

4.2.11 Sensitivity Analysis: ............................................................................................. 75

4.3 Results ............................................................................................................................... 76

4.3.1 Base-case: ............................................................................................................. 76

4.3.2 Sensitivity Analysis: ............................................................................................. 77

4.4 Discussion ......................................................................................................................... 79

5 Synthesis .................................................................................................................................. 95

5.1 Novel Findings .................................................................................................................. 95

5.2 Implications for Clinical Practice ..................................................................................... 96

5.3 Foci for Future Research ................................................................................................... 98

5.3.1 Real World Outcomes ........................................................................................... 98

5.3.2 Utilities .................................................................................................................. 99

5.3.3 Mortality is the Key Parameter ............................................................................. 99

5.4 Conclusion ...................................................................................................................... 100

References ................................................................................................................................... 101

Page 8: Economic Evaluation of Percutaneous Coronary Intervention in

viii

List of Tables

page

Chapter 1

Table 1: Characteristics of Economic Evaluations of Medical Therapy vs. PCI…………. 11

Chapter 2

Table 1: Trial Design and Characteristics of Randomized Controlled Trials of Medical…

Therapy versus PCI

28

Table 2: Baseline Medication Use in the Medical Therapy and the PCI Groups………… 31

Table 3: Study Quality and End Point Ascertainment in Randomized Controlled Trials…

of Medical Therapy versus PCI

33

Chapter 3

Table 1: Model Specification…………………………………………………………….. 57

Table 2: Baseline Characteristics of Study Cohort……………………………………….. 58

Table 3: Results of Model Derivation……………………………………………………. 59

Table 4: Model Performance……………………………………………………………… 60

Chapter 4

Table 1: Probabilities……………………………………………………………………... 84

Table 2: Hazard Ratios……………………………………………………………………. 85

Table 3: Utilities…………………………………………………………………………... 86

Table 4: Costs……………………………………………………………………………... 87

Table 5: Validation………………………………………………………………………... 88

Table 6: Life Time Cumulative Probability of Myocardial Infarction & PCI……………. 89

Table 7: Base Case Results……………………………………………………………….. 90

Table 8: Sub-group Analysis based on Risk of Restenosis……………………………….. 91

Table 9: Scenario Analysis based on Initial Symptom Severity………………………….. 92

Page 9: Economic Evaluation of Percutaneous Coronary Intervention in

ix

List of Figures

page

Chapter 1

Figure 1: Study Selection…………………………………………………………………. 12

Chapter 2

Figure 1: Process of Study Selection……………………………………………………... 35

Figure 2: Summary Odds Ratio of Freedom from Angina……………………………….. 36

Figure 3a: Summary Odds Ratio Stratified by Duration of Follow-Up…………………... 37

Figure 3b: Summary Odds Ratio Stratified by Period of Study Recruitment…………….. 37

Figure 4: Meta-Regression of Freedom of Angina on Evidence Based Medications…….. 38

Chapter 3

Figure 1: Predicted versus Observed EQ-5D scores in the Validation Dataset………….. 62

Chapter 4

Figure 1: Model Structure………………………………………………………………. 93

Figure 2: Cost-Effectiveness Acceptability Curve………………………………………... 94

Page 10: Economic Evaluation of Percutaneous Coronary Intervention in

x

List of Appendices

page

Chapter 1

Appendix A: Search Strategy……………………………………………………………... 128

Chapter 2

Appendix A: Analysis restricted to non-myocardial infarction trials………………... 130

Appendix B: Meta-regression, restricted to non-myocardial infarction trials……………. 130

Chapter 3

Appendix A: Multiple Imputation Analysis……………………………………………… 131

Appendix B: Winbugs Code……………………………………………………………… 132

Appendix C: Subgroup Analysis for Mean Prediction in Validation dataset……………. 133

Chapter 4

Appendix A: Medical Therapy Sub-tree………………………………………………….. 134

Appendix B: PCI Sub-tree………………………………………………………………... 135

Appendix C: MI Sub-tree…………………………………………………………………. 136

Appendix D: CABG Sub-tree…………………………………………………………….. 137

Appendix E: Validation of Subgroup Analyses…………………………………………... 138

Page 11: Economic Evaluation of Percutaneous Coronary Intervention in

1

1 Introduction

Cardiovascular disease remains the most prominent cause of death in industrialized countries,

and represents a substantial economic burden, accounting for 18% of the overall health care

system costs in Canada (1). The burden of cardiovascular disease is predominantly in the

ambulatory setting, specifically in patients with stable coronary artery disease, with a

substantially smaller proportion hospitalized with acute coronary syndromes or myocardial

infarctions (MI) (2). As such, advances in the therapeutic options for patients with stable

coronary artery disease have substantial impact on overall cardiovascular mortality (2). In

contemporary practice, the alternative treatment options for a patient with symptomatic stable

coronary disease are medical therapy alone, or in combination with revascularization with either

coronary artery bypass grafting (CABG) or percutaneous coronary intervention (PCI) (3).

CABG is generally restricted to patients with severe multi-vessel coronary artery disease while

the revascularization modality of choice for more focal coronary disease is PCI (4).

1.1 Contemporary Treatment Options: PCI or Medical Therapy

PCI has figured prominently in the contemporary treatment of patients with coronary artery

disease. In the province of Ontario, Canada approximately 21,000 PCIs are performed annually,

with more than 1.3 million annual procedures being performed in the United States (5;6). Since

its advent in 1979, PCI has been highly efficacious in the treatment of angina, by restoring

luminal patency by balloon inflation at the site of the luminal stenosis in the epicardial coronary

artery. Over the last two decades, PCI has evolved to include the placement of intra-coronary

bare metal stents (BMS) which act as scaffolding to prevent arterial recoil. Despite its short term

efficacy, long term results with BMS have been limited by accelerated neo-intimal proliferation

at the site of the stent placement, resulting in luminal restenosis (3). Restenosis is often

accompanied by the recurrence of anginal symptoms and typically requires repeat PCI, or

occasionally surgical revascularization with CABG.

To address this “Achilles heel” of BMS, stents coated with anti-proliferative medications were

developed (3). These drug-eluting stents (DES) slow or inhibit neo-intimal proliferation by the

controlled luminal release of the anti-proliferative medications. In 2002, the first generation of

polymer-based DES, the Sirolimus-eluting CYPHER stent (Johnson & Johnson) and the

Page 12: Economic Evaluation of Percutaneous Coronary Intervention in

2

Paclitaxel-eluting TAXUS stent (Boston Scientific) were approved for use in Canada. Since

then, advances in the design of the stent platforms, associated polymers and anti-proliferative

medications has led to the introduction of 2nd

and 3rd

generation DES, which are now available

for use.

All DES markedly reduce restenosis and recurrent angina when compared to BMS (7).

However, concerns regarding their long-term safety have been raised by a number of studies

which have suggested a potential increase in late myocardial infarctions (MI) from very late stent

thrombosis due to incomplete endothelization of stent struts (7-10). Although late stent

thrombosis is rare, with estimates suggesting a 0.13% increased risk per year compared to BMS,

it has an extremely poor prognosis, with short term mortality estimated at 20% (10). In

comparison to BMS, DES stents are marginally less deliverable and have substantially higher

acquisition costs. Moreover, prolonged administration of dual anti-platelet coverage until

adequate stent strut endothelization, with a combination of aspirin and a thienopyridine such as

Clopidogrel, is critical to reduce the risk of stent thrombosis (11). Newer thienopyridines are

expensive and dual anti-platelet coverage has a non-trivial risk of major bleeding (12).

Concurrent to advances in PCI technology, baseline medical therapy has improved dramatically,

with the introduction of angiotensin-converting enzyme (ACE) inhibitors and HMG-CoA

reductase inhibitors (statins). The Clinical Outcomes Utilizing Revascularization and

Aggressive Drug Evaluation (COURAGE) trial compared an initial strategy of optimal medical

therapy versus PCI with BMS in patients with stable coronary artery disease, and found no

survival difference (13). This finding has been confirmed by several meta-analyses focusing on

death, myocardial infarction, and repeat coronary revascularization, all consistently

demonstrating no incremental benefit of PCI compared with medical therapy in reducing the risk

of death or MI (14;15).

1.2 Evaluation of Angina Relief Associated with PCI

Accordingly, relief of angina may be the most appropriate indication for PCI among patients

with stable coronary artery disease in contemporary interventional practice; yet, no systematic

review has estimated the efficacy of PCI for angina relief compared with medical therapy.

Although evidence from early randomized trials has shown that PCI provides substantial angina

relief compared with medical therapy, more recently published trials have challenged this

Page 13: Economic Evaluation of Percutaneous Coronary Intervention in

3

conventional wisdom (16-20). For example, in the Randomized Intervention Treatment of

Angina (RITA) 2 trial, investigators found that PCI patients were almost twice as likely to be

angina free at 3 years (16;21). In contrast, in the COURAGE study, investigators only found a

small early incremental benefit associated with PCI for angina relief compared with optimal

medical therapy (20;22). These discrepancies underscore the need to systematically evaluate the

efficacy of PCI versus medical therapy for long term angina relief.

Meta-analytical techniques are the most common statistical tool to summarize the evidence from

independent studies and provide a single summary estimate of treatment effect, which in this

case would be angina relief (23). A fundamental principal of meta-analyses is that the same

research question is being addressed on a similar patient population, such that statistical pooling

from separate studies is valid (24). Several features of this clinical condition make this

problematic.

First, there is no census on the definition of stable angina or chronic stable coronary artery

disease. Coronary artery disease represents a spectrum that ranges from asymptomatic patients

with non-obstructing lesions, to patients with stable angina with flow-limiting lesions causing

ischemia, to patients with unstable acute coronary syndromes where disruption of vulnerable or

high-risk plaques act as a stimulus for thrombogenesis (24;25). The clear delineation of when a

patient with an acute coronary syndrome becomes stable is potentially difficult as patients may

undergo this transition rapidly. Moreover, within the „stable‟ population, even patients with

flow-limiting lesions, can range from asymptomatic to severely symptomatic with chest pain

with minimal exertion.

The distinction between stable versus unstable patients is important in evaluating the possible

benefit of PCI as it has been consistently demonstrated to be beneficial compared with medical

therapy among patients with acute coronary syndrome. Systematic reviews of randomized trials

evaluating primary PCI for patients with ST-segment elevation myocardial infarction (STEMI)

have demonstrated absolute reductions of 2% for mortality and 4% for myocardial infarction

compared with fibrinolytic therapy (26). Even among STEMI patients who were successfully

treated with fibrinolytic therapy, recent evidence has suggested that PCI is associated with a

mortality reduction comparable with medical therapy (27). Similarly, several landmark studies

have demonstrated the benefit of PCI in reducing death and recurrent MI among patients with

Page 14: Economic Evaluation of Percutaneous Coronary Intervention in

4

non-ST elevation myocardial infarction (NSTEMI) (25). Therefore, systematic reviews

evaluating stable coronary artery disease that include a relatively high proportion of patients who

have recently recovered from an acute coronary syndrome would likely overestimate the benefit

of PCI, when compared an analysis restricted to a more stable population. Despite this, many

trials that were designed to evaluate a stable angina population often recruited such patients.

Even in the COURAGE trial, the landmark trial in this area, enrollment criteria included patients

who were stable after myocardial infarction (without specifying the duration of stabilization) in

addition to patients with symptomatic chronic angina pectoris and asymptomatic patients with

objective evidence of myocardial ischemia (20).

Second, as mentioned earlier, PCI and medical therapy have both progressed tremendously over

the last two decades. As such, studies completed earlier in this time period may not be

comparable to more contemporary studies. For example, in the RITA-2 study which recruited

patients from 1992-1996, coronary stent use was limited to 9% of PCI patients, and less than

20% of enrolled patients were prescribed an ACE-inhibitor or a statin medication (16;21). In

contrast, in the COURAGE trial (which recruited patients from 1999 to 2004), 94% of PCI

patients received stents, and 58% and 89% were on ACE-inhibitors and statins respectively (20).

Thus pooling studies from different eras may be inappropriate. Comparing PCI versus medical

therapy in these different time periods may provide insight as to relative impact of advancements

in medical therapy versus technological progress in PCI on the prognosis of patients.

Finally, the endpoint of angina relief does not have a common definition. The Canadian

Cardiovascular Society (CCS) functional class is the most widely accepted scale for measuring

angina severity. Thus, the most appropriate measure of angina relief may be change in angina

severity based on CCS class over the course of the study. However, this is selectively reported

in the clinical trials.

Even in the absence of suchlimitations, all meta-analyses typically have some degree of

heterogeneity (23). The presence of heterogeneity in meta-analysis is frequently assessed by the

Cochran Q-statistic, and the proportion of variability due to heterogeneity between individual

trials is frequently quantified by the I2 index (24;28). Newer techniques that explore

heterogeneity include subgroup analysis and meta-regression techniques (23). The thorough

Page 15: Economic Evaluation of Percutaneous Coronary Intervention in

5

consideration of heterogeneity using these techniques is of particular relevance to angina relief in

stable coronary disease, so as to provide insights as to the divergent findings in the literature.

1.3 Impact of Angina on Quality of Life

Importantly, as the decision to perform a PCI is based on the efficacy of either a BMS or DES in

relieving symptoms, as compared to medical therapy alone, it is imperative to have accurate

estimates of the impact of angina on quality of life. Health status measurement instruments

provide a numerical score representing health profile across health domains such as physical,

mental, and social well-being (29-31). Such instruments can be classified into four categories

based on the following criteria: 1. whether the instrument is designed for a generic versus a

disease-specific population, and 2. whether it is a psychometric/descriptive versus a preference-

based instrument (32). Both descriptive and preference based instruments provide a score

reflective of a patients‟ assessment of their own quality of life. The preference-based measures

add a valuation component to the patient report. For example the EQ-5D, a self-administered

questionnaire consisting of five health dimensions (mobility, self-care, usual activities, pain, and

mood) uses community-based preferences as their valuation component (33;34).

The preference for a health state is a quantitative measure of the desirability or undesirability of

that particular health state (35). It is represented by a utility weight that ranges between a lower

anchor of zero, which represents dead, and an upper anchor of one, which represents perfect

health. Where a particular health state exists in this continuum can be determined by direct

utility elicitation methods such as the standard gamble or time trade off. These involve patient

interviews, are very labor intensive, and can be difficult for patients to comprehend (35).

Indirect utility elicitation is made possible by preference based instruments such as the EQ-5D

(36). After a patient scores their current health state on the EQ-5D, the utility weight of that

particular health state is determined by applying a country specific tariff (33;36). The tariffs in

turn were developed in independent studies, where a subset of the possible EQ-5D health states

were directly valuated by the time trade off method(33).

The most common instrument for quantifying the impact of angina on quality of life is the

Seattle Angina Questionnaire (SAQ) (37;38). This is a disease-specific psychometric/descriptive

tool that provides a score reflective of a patient‟s assessment of their own quality of life in the

context of symptoms and impairments unique to coronary artery disease (37;38). It is widely

Page 16: Economic Evaluation of Percutaneous Coronary Intervention in

6

used and has been shown to be valid, responsive and sensitive to changes in angina severity

(37;38). However, the SAQ, as a psychometric/descriptive tool cannot be used to calculate

quality–adjusted life years (QALY‟s), the health index of choice for economic evaluations (3).

The calculation of the QALY requires a valuation component that reflects community-based

preferences for a particular health state, such as angina – this requires a preference based

instrument such as the EQ-5D. Preference based instruments are rarely included in

cardiovascular clinical trials or outcomes research (39).

Quality-adjusted survival is of particular relevance to chronic stable angina, given that the key

distinction between the alternative treatment options of PCI and medical therapy alone is

hypothesized to be differences in symptom control and therefore quality of life. The paucity of

comprehensive and current data with which to estimate the quality of life component of quality

adjusted survival in patients with stable coronary artery disease is a potential limitation to the

validity and credibility of economic evaluations in this area (39). Current methodological

guidelines for economic evaluations recommend the use of utility weights directly measured

using the EQ-5D; in the absence of this, a validated mapping algorithm should be used (40).

Mapping algorithms have been developed to allow the prediction of utility weights from the

scores on a descriptive instrument (32). The most common technique for mapping is called the

„transfer to utility regression method‟; in this method, a regression model is created with the

summary score from the utility-based target as the dependent variable and the individual

components of the descriptive measure as the co-variates of interest (32). Other mapping

techniques include direct revaluation and effect size translation, both of which involve direct

utility measurement using standard gamble or time trade off (32). Finally, response mapping

techniques have been described to predict an individual patient‟s responses on each item of

preference based instrument, from responses on the descriptive instrument (32). The appeal of

the transfer to utility method lays in its simplicity and minimal data requirements. This approach

has been the technique of choice in mapping algorithms, accounting for approximately 70% of

published studies (32).

The conversion of SAQ scores to utility weights has not previously been attempted (32). We

believe that this will be of substantial use to health services researchers by making data from

previous studies which have measured quality of life using descriptive instruments in stable

Page 17: Economic Evaluation of Percutaneous Coronary Intervention in

7

coronary artery disease available for secondary use in economic analyses, to calculate quality-

adjusted survival.

1.4 Evaluating trade-offs: a decision analytic framework

From a health policy perspective, it remains unclear if the degree to which PCI with either BMS

or DES reduces angina and repeat procedures in comparison to medical therapy alone, offsets the

increased costs and the potential safety risks. This decision requires balancing clinical efficacy

and safety, incorporating the uncertainties surrounding these estimates, while simultaneously

taking into account patient values and also minimizing resource utilization and costs (41). A

decision analytic framework is the ideal tool for evaluating such trade-offs and potentially

clarifying this clinical controversy (42;43). A decision model simulates transitions, based on

probabilities, between potential future outcomes following the alternative initial strategies being

evaluated (42;43). Each outcome has an associated value, which can be expressed as cost in

terms of resource use, or as a clinical consequence, expressed as the life years or quality adjusted

life-years gained (42;43). The expected cost and consequences of each alternative strategy is the

sum of costs and consequences associated with future stream of outcomes, weighted by the

probability of each outcome. Depending on whether only clinical outcomes, or if both costs and

consequences are considered, a decision analytic model can be used as a clinical decision making

tool, or for full health economic evaluations respectively.

Key elements when developing a decision model are first to consider all relevant alternative

strategies for the clinical condition being evaluated (44-47). For patients with stable coronary

artery disease, the relevant treatments strategies are medical therapy with or without

revascularization. If the study population excludes patients with severe multi-vessel disease, the

revascularization modalities of relevance are PCI with either BMS or DES.

Having determined all the relevant alternatives, the consequences of each must be modelled,

over a time horizon that will capture all potential differences between strategies (44-47). The

time horizon represents the time frame over which consequences will be modelled. In order to

capture all relevant effects, it is generally recommended that a life-time time horizon is adopted.

The next step is to integrate all available evidence to populate the model (44-47).

Page 18: Economic Evaluation of Percutaneous Coronary Intervention in

8

Rarely will all the relevant evidence describing the effectiveness, resource use and quality of life

weights of an intervention come from single source (42;43). The advantage of the decision

analytic framework is that it allows one to draw on the best evidence from multiple sources,

including randomized clinical trials, and observational studies, such as cohort studies, or surveys.

Through the structure of the model, which must reflect all the relevant future health states of a

particular clinical condition, evidence from these different sources can be incorporated, and

translated into estimates of costs and consequences (42). Necessarily, the evidence used to

populate a model will have varying degrees of quality, and as such, there will be varying degrees

of uncertainty associated with each parameter. Importantly, a decision model provides a

framework for decision-making under conditions of uncertainty, and a mechanism for

understanding the source of this uncertainty, so as to target future research (42).

1.5 Literature to date

To understand the degree to which these key points have been addressed, we performed a

systematic review of the published literature on economic evaluations in chronic stable angina.

The specifics of the search keywords can be found in Appendix 1. Briefly, relevant published

studies were identified through a computerized literature search of the MEDLINE and EMBASE

electronic databases from January 1950 to February 2011, using the terms “transluminal

percutaneous coronary angioplasty”, and “angioplasty” and a search filter for economic analyses.

OVID search software was used, with the „exploded‟ search feature. In addition, bibliographies

of journal articles and relevant reviews were extensively hand-searched to locate additional

studies. Relevance for inclusion in the systematic review was determined using a hierarchical

approach based on title, abstract, and the published manuscript. The search results are outlined

in Figure 1.

We found 1816 citations, of which 193 abstracts were reviewed after a title screen. 37 full text

published manuscripts were evaluated, of which we found only four relevant studies (48-51).

The reasons for exclusion are detailed in Figure 1. There were 13 review articles or editorials

(52-64). Two studies compared 2nd

generation DES to 1st generation DES (65;66), and 15

compared DES to BMS only (67-81). Of the excluded studies which had a medical therapy arm,

one was a decision model comparing medical therapy to balloon angioplasty (82), while two

were cost analyses only (83;84).

Page 19: Economic Evaluation of Percutaneous Coronary Intervention in

9

The characteristics of the four included studies are found in Table 1. The Trial of Invasive

versus Medical Therapy in the Elderly (TIME) suggested that PCI was cost-effective over a time

horizon of 1 year, but did not assess QALYs (48). Instead, this trial focused on the costs to

prevent a major event, defined as death, non-fatal MI, or hospitalization for uncontrolled

symptoms (48). The analysis from the COURAGE trial showed that initial PCI with BMS was

not cost-effective, with an ICER of $168,000 per QALY gained (48;51). The third trial, an

economic analysis of the Bypass Angioplasty Revascularization 2 Diabetes (BARI2D) study,

included only diabetic patients, and again concluded that PCI (24% DES use) was not cost-

effective over a 4 year period (50). The final study by Griffin and colleagues was a retrospective

observational analysis, comparing the cost and quality-adjusted survival over 6 years in 385

patients deemed appropriate for PCI, categorized based on initial medical versus PCI versus

surgical treatment strategy. This analysis also found that PCI was not cost-effective.

Several limitations of these previous analyses merit discussion. First, were all relevant

alternatives considered? Of the four analyses, only one considered the three relevant strategies

of BMS, DES and medical therapy, and this was restricted to a diabetic population. The time

horizon of the studies was generally short. Only one of the studies had a life-time time horizon

in its primary analysis, which is critical in order to capture any relevant mortality effects (51).

Given the importance of symptom relief in the decision between medical therapy alone versus

PCI, the valuation of angina is key. In the two studies where QALY was the metric of choice,

different scaling methods were used. The study by Griffin et al. used utilities from the EQ-5D

tool, while utilities in the COURAGE trial were measured directly from the patients using

standard gamble (49;51;57;85). It is well recognized that different scaling methods for utility

elicitation may result in different weights for the same health state, making comparisons across

these studies problematic (29;86;87).

Finally, was all available best evidence used to populate the analyses? Of the four previous full

economic analyses comparing medical therapy to invasive therapy with stenting, three were

randomized trials (48-51). Although randomized clinical trials provide unbiased estimates of

efficacy, they are limited by highly restrictive enrolment criteria. For example, in the

COURAGE trial, only 6.4% of screened patients with coronary disease were enrolled in the trial

(13). Moreover, the populations in the studies were selective, with some enrolling only

diabetics, while others including only the elderly (48;50). The variation in initial symptom

Page 20: Economic Evaluation of Percutaneous Coronary Intervention in

10

severity across studies was marked, reflecting the inclusion criteria of the trials. As such, both

costs and consequences from these studies are not generalizable to real world practice.

1.6 Gaps in Knowledge

Based on this review, we believe there are three major gaps in the knowledge which must be

addressed. First, the degree of angina relief associated with PCI as compared with medical

therapy in patients with stable coronary artery disease must be estimated through a systematic

review of the literature. We anticipate that there will be substantial heterogeneity across trials,

and elucidation of this heterogeneity will provide insight as to the contradictory findings in the

literature. Second, more detailed and up-to-date data on utility weights for varying angina

severity are required. A conversion/mapping tool to make SAQ data from previous studies

available for secondary use in cost-effectiveness analyses will be of substantial use to health

services researchers, allowing for valid comparisons across angina severity and between studies.

Finally a comprehensive decision analytic model which evaluates the incremental cost-

effectiveness of PCI with either DES or BMS compared to aggressive medical therapy alone for

patients with stable coronary artery disease, calibrated to real world data, and incorporating all

sources of previous evidence, will address the limitations of previous studies in this area. This in

turn, will inform policy decisions as to the most appropriate initial strategy in this large

population of patients.

This thesis will address each of these gaps in the following three chapters, each of which is a

manuscript (published or in-submission). The final synthesis chapter will summarize the novel

findings of the thesis, and provide foci for further work.

Page 21: Economic Evaluation of Percutaneous Coronary Intervention in

11

Table 1: Characteristics of Economic Evaluations of Medical Therapy vs. PCI

Study Design Strategies Country n Population Outcomes Time

Horizon Conclusion

Inclusion Gender Symptoms Effectiveness Monetary

TIME,

2004 RCT

Medical vs.

PCI Sweden 188

> 75 years

old only

44%

women

75% CCS

3-4

Major event

averted

Swiss

Currency 1 year

PCI is cost

effective in

averting

major events

( €6965 per

event

averted)

Griffin et

al, 2007 Cohort

Medical vs.

PCI England 385

Stable

angina NA NA QALY (EQ5D) £ 6 years

47,000

£/QALY

gained for

PCI

COURAGE

, 2008 RCT

Medical vs.

BMS

Multi-

national 2287

Stable

angina

15%

women 23% CCS 3

QALY

(standard

gamble)

2004 US $ Life-time

ICER for

PCI

$168,000 -

$300,000

BARI2D,

2009 RCT

Medical vs.

DES/BMS

Multi-

national 1347

Diabetic

only

32.2%

women

7.9% CCS

3-4

10.7% with

ACS

Life-years 2007 US $ Life-time

Medical

therapy was

dominant

RCT: randomized control trial; DES: drug eluting stent; BMS: bare metal stent; CCS: Canadian cardiovascular society class; ACS: acute coronary syndrome; QALY:

quality adjusted life-year; PCI: percutaneous coronary intervention; ICER: incremental cost-effectiveness ratio

Page 22: Economic Evaluation of Percutaneous Coronary Intervention in

12

Figure 1: Study Selection

Page 23: Economic Evaluation of Percutaneous Coronary Intervention in

13

2 Effects of PCI versus Medical Therapy on Angina Relief: A Meta-analysis

Authors:

Harindra C. Wijeysundera MD1, Brahmajee K. Nallamothu MD, MPH

2, Harlan M. Krumholz

MD, SM3, Jack V. Tu MD, PhD

1,4, 5, Dennis T. Ko MD, MSc

1, 4, 5

Affiliations:

1Division of Cardiology, Schulich Heart Centre and Department of Medicine, Sunnybrook

Health Sciences Centre, University of Toronto, Ontario, Canada

2Health Services and Research Development Center of Excellence, Ann Arbor VA Medical

Center, Ann Arbor, Michigan

3Section of Health Policy and Administration, Department of Epidemiology and Public Health;

Robert Wood Johnson Clinical Scholars Program, Yale University School of Medicine; and the

Center for Outcomes Research and Evaluation, Yale-New Haven Health, New Haven,

Connecticut

4Institute for Clinical Evaluative Sciences, Toronto, Ontario, Canada

5Department of Health Policy, Management and Evaluation, University of Toronto, Ontario,

Canada

Contents of this chapter have been published in the Annals of Internal Medicine: Harindra C.

Wijeysundera, et al, Meta-analysis: Effects of Percutaneous Coronary Intervention Versus

Medical Therapy on Angina Relief. Ann Intern Med March 16, 2010 152:370-379

Page 24: Economic Evaluation of Percutaneous Coronary Intervention in

14

ABSTRACT

Background

Several meta-analyses have evaluated the efficacy of percutaneous coronary intervention (PCI)

compared with medical therapy, but none has focused on angina relief.

Purpose

To summarize evidence on the degree of angina relief provided by PCI compared with medical

therapy in patients with stable coronary artery disease.

Data Sources

The Cochrane library (1993 – June, 2009), EMBASE (1980 - June, 2009) and MEDLINE (1950

- June, 2009) electronic databases, with no language restrictions.

Study Selection

Two independent reviewers screened citations to identify randomized controlled trials of PCI

versus medical therapy in patients with stable coronary artery disease.

Data Extraction

Data on patient characteristics, study conduct and outcomes were abstracted by two independent

reviewers. A random-effects model was used to combine data on freedom from angina and to

Page 25: Economic Evaluation of Percutaneous Coronary Intervention in

15

perform stratified analyses based on duration of follow-up, post-myocardial infarction status,

coronary stent utilization, period of recruitment, and utilization of evidence-based medications.

Data Synthesis

A total of 14 trials, enrolling 7818 patients met our inclusion criteria. Although PCI was

associated with an overall benefit on angina relief (odds ratio [OR] 1.69; 95% Confidence

Interval [CI] 1.24-2.30), important heterogeneity across trials was observed. The incremental

benefit of PCI observed in older trials (OR 3.38; 95% CI 1.89-6.04) was substantially less and

possibly absent in recent trials (OR 1.13; 95% CI 0.76-1.68). An inverse relationship between

utilization of evidence-based therapies and the incremental benefit of PCI was observed.

Limitations

Information about long term medication use was incomplete in most trials. Few trials used drug

eluting stents. Meta-regression analyses used aggregated study-level data from few trials.

Conclusions

PCI was associated with greater freedom from angina compared with medical therapy but this

benefit was largely attenuated in contemporary studies. This observation may be related to

greater use of evidence-based medications.

Page 26: Economic Evaluation of Percutaneous Coronary Intervention in

16

2.1 Introduction

Percutaneous Coronary Intervention (PCI) has figured prominently in the treatment of patients

with stable coronary artery disease. It is estimated that more than 400,000 PCIs are performed

for this indication each year in the United States (88). Several meta-analyses have compared the

efficacy of PCI versus medical therapy in patients with stable coronary disease; however, these

analyses have focused on death, myocardial infarction, and repeat coronary revascularization and

have consistently demonstrated no incremental benefit of PCI compared with medical therapy

(14;15). Accordingly, relief of angina may be the most appropriate indication for PCI among

patients with stable coronary artery disease in contemporary interventional practice; yet, no

systematic review has estimated the efficacy of PCI for angina relief compared with medical

therapy.

Although evidence from early randomized trials has shown that PCI provides substantial angina

relief compared with medical therapy, more recently published trials have challenged this

conventional wisdom (16-20). For example, in the Randomized Intervention Treatment of

Angina (RITA) 2 trial, published in 1997, investigators found that PCI patients were almost

twice as likely to be angina free at 3 years (16;21). In contrast, in the Clinical Outcomes

Utilizing Revascularization and Aggressive Drug Evaluation (COURAGE) study, published in

2007, investigators only found a small early incremental benefit associated with PCI for angina

relief compared with optimal medical therapy (20).

These discrepancies underscore the need to systematically evaluate the efficacy of PCI versus

medical therapy for long-term angina relief. Accordingly, we performed a systematic review of

the literature to best estimate the degree of angina relief associated with PCI as compared with

medical therapy in patients with stable coronary artery disease. Using pre-specified subgroup

analysis and meta-regression techniques, we also evaluated several key study-level factors that

might have contributed to heterogeneity across trials, such as the length of follow-up, inclusion

Page 27: Economic Evaluation of Percutaneous Coronary Intervention in

17

of patients with recent acute myocardial infarction, utilization of coronary stents, period of study

recruitment, and utilization of evidence-based therapies.

2.2 Methods

A standard protocol for study identification, inclusion and data abstraction was developed and

adhered to for all steps of the systematic review. All subgroup and meta-regression analyses

outlined below were pre-specified in this protocol.

2.2.1 Data Sources and Searches

Relevant published studies were identified through a computerized literature search of the

Cochrane library (1993 to June, 2009), EMBASE (1980 to June, 2009) and MEDLINE (1950 to

June, 2009) electronic databases, using the terms transluminal percutaneous coronary

angioplasty, and angina pectoris. OVID search software was used with the „exploded‟ search

feature, and searches included both English and non-English language records (89). Two

independent reviewers screened citations for inclusion in the systematic review using a hierarchical

approach, first assessing the title, then the abstract, and the published manuscript (HW and DK). In

addition, bibliographies of journal articles and relevant reviews were extensively hand-searched

to locate additional studies (HW).

2.2.2 Study Selection

Randomized trials were included if they enrolled patients with stable, coronary artery disease and

compared a treatment strategy of PCI versus medical therapy. Since there is no universally-

acceptable definition of when a patient with unstable coronary artery disease becomes stable, we

included trials that enrolled patients with recent acute coronary syndromes that had been

stabilized for more than a week (19;90-94). We also included trials that enrolled patients with

minimal or no angina with stable coronary artery disease. Inclusion of these trials in our study is

similar to previous meta-analyses evaluating patients with stable coronary artery disease

(14;15;95).

Page 28: Economic Evaluation of Percutaneous Coronary Intervention in

18

2.2.3 Data Extraction

Data on patient characteristics, study conduct and outcomes were abstracted by two independent

reviewers (HW and DK). The primary outcome was freedom from angina, which was abstracted

using the original definition from published manuscripts. Where this was not available, freedom

from angina was estimated using surrogate information (Table 3). Although the assessment of

freedom of angina across the trials was not standardized, within each trial the same criteria were

applied equally to the treatment groups. In our main analysis, we assessed events from the

longest follow-up period available in the published articles. We used the number of patients for

whom symptom assessments were available at the follow-up period as the denominator in our

calculations when determining the proportion of patients who were free from angina. This

would minimize bias in our meta-analysis by ensuring the weight of a particular study was

reflective of the number of patients in whom assessments were actually performed.

We further stratified our analysis based on the duration of follow-up (less than 1 year, between 1

to 5 years and greater than 5 years). For studies that reported outcomes for multiple follow-up

periods, we included the trial data in all applicable subgroups. We restricted our review to the

overall results of the trial and did not include selective data from sub-group analyses.

Study quality was evaluated based on the 5-point scale outlined by Jadad et al, with criteria for:

randomization with proper concealment of the allocation sequence, blinding of the patient and

investigator to treatment allocation with description of the blinding method, and a description of

withdrawals and dropouts (96).

2.2.4 Data Synthesis and Analysis

We explored heterogeneity between trials using both subgroup analysis and meta-regression

techniques. First, we performed a subgroup analysis stratifying trials based on recent myocardial

infarction status. Recent myocardial infarction trials were defined as studies with enrollment

restricted to patients who were stable at least one week post myocardial infarction. Second, we

Page 29: Economic Evaluation of Percutaneous Coronary Intervention in

19

performed an analysis based on whether more than 50% of the PCI procedures in a trial included

coronary stent implantation. Third, we stratified trials into three 5-year periods based on the last

year of patient recruitment: for older trials recruitment ended from 1990-1994, for intermediate

trials recruitment ended from 1995-1999, and for contemporary trials recruitment ended in 2000

and later.

Finally, we performed a univariate meta-regression analysis with the number of evidence-based

therapies as the predictor of interest. Baseline utilization rates of evidence-based therapies

including aspirin, β-blocker, angiotensin converting enzyme (ACE)-inhibitor, and statin were

ascertained in each trial. The use of other anti-anginal medications such as long acting nitrates

and calcium-channel blockers was inconsistently reported in the trials; hence these medication

classes were not explored. Aspirin, β-blocker, statin and ACE-inhibitor use was evaluated given

their strong benefits for patients with coronary artery disease and thus their utilization was likely

a proxy for the aggressiveness of overall medical therapy in trials. A threshold of 50% was used

to categorize the utilization of each medication in each trial. Additional sensitivity analyses were

performed using different utilization thresholds. We contacted the study authors as needed

regarding data not available in the published manuscripts (90;93;97;98).

A random-effects model based on the DerSimonian and Laird method for combining results from

individual trials was used. Summary odds ratio (OR) and 95% confidence intervals (CI) were

calculated. Heterogeneity was evaluated by calculating the Cochran Q-statistic, which

determines if statistically significant heterogeneity was present, and the I2 index, which describes

the proportion of variability due to heterogeneity between individual trials (99). Due to the

likelihood of heterogeneity over time periods and across patient types, a random effects model

was used as it incorporates both within trial and between trial variance.

Subgroup analyses were performed using random-effects models restricted to the trials within the

subgroup of interest. We developed a random-effects meta-regression model by plotting the

natural logarithm of the odds ratio for freedom from angina against the number of evidence

Page 30: Economic Evaluation of Percutaneous Coronary Intervention in

20

based medications for which there was greater than 50% utilization (100-103). This model

extends the DerSimonian and Laird meta-analytical technique by incorporating co-variates in

order to reduce the residual heterogeneity (100-103).

Sensitivity analyses were conducted to examine the robustness of our results by eliminating one

study at a time from the analyses to determine if the pooled estimates were disproportionately

influenced by a particular trial. We found that both our overall random-effects model and meta-

regression model were robust.

Publication bias was assessed qualitatively using a funnel plot, where the treatment effect of each

included study was plotted against the study size and quantitatively with the Egger‟s Test of the

intercept. The funnel plot was symmetric and the Egger‟s test was not significant, both

suggesting against publication bias.

Statistical significance was set as a p-value less than 0.05. All statistical calculations were

performed with the use of Comprehensive Meta-analysis Version 2 Software (Biostat,

Englewood NJ 2005).

2.2.5 Role of the Funding Source

This study was funded in part by operating grants by a Canadian Institutes of Health Research

(CIHR) (MOP 82747) and a CIHR Team Grant in Cardiovascular Outcomes Research. All

decisions regarding study design, data analysis, and publication were made independent of the

funding agency.

Page 31: Economic Evaluation of Percutaneous Coronary Intervention in

21

2.3 Results

2.3.1 Study Selection

Of the 310 citations reviewed, 22 articles were retrieved that met our inclusion criteria (16-

21;90-94;97;98;104-117) (Figure 1, available at www.annals.org). We excluded eight of these

trials, four that did not report on angina (113-115;117), and four trials that reported outcomes of

overall coronary revascularization but did not distinguish between PCI and coronary artery

bypass grafting surgery outcomes (104;110-112). Our meta-analysis was based on the remaining

14 randomized trials, enrolling a total of 7818 patients.

2.3.2 Study Design

Table 1 summarizes the design characteristics of the 14 remaining trials. The majority of trial

patients were male, with normal left ventricular systolic function. The proportion of patients

with diabetes mellitus ranged from 9% in the RITA-2 trial to 33% in the COURAGE trial

(20;21); 57.5% had single vessel disease, 28.8% had double vessel disease and 13.6% had triple

vessel coronary artery disease. Six trials, with 3010 patients, restricted enrollment to patients

who had been stabilized following an acute myocardial infarction (19;90-94).

Coronary stent use varied significantly among trials included in the meta-analysis. While no

patients received stents in the Medicine, Angioplasty or Surgery (MASS) study published in

1995, the majority of the patients enrolled in the later trials did (7,8). Only the Occluded Artery

Trial (OAT) and COURAGE trials reported drug eluting stent use, at 8% and 2.6% respectively

(19;20). As shown in Table 1, there was a significant proportion of patients with minimal or no

symptoms (i.e. Canadian Cardiovascular Society functional classification 0 or 1); this was

especially the case in the six post-myocardial infarction trials (7,11-15).

As illustrated in Table 2, there were significant differences in baseline utilization rates for

evidence-based medications among trials. Although aspirin was used frequently in all studies

with rates ranging from 75% to 100%, utilization of other medications varied widely among

Page 32: Economic Evaluation of Percutaneous Coronary Intervention in

22

trials. For example, ACE-inhibitor use ranged from 8% to 88%, while statin use ranged from

12% to almost 90% in more recent trials. Rates of medication use were similar between patients

randomized to PCI versus medical therapy within each trial with rare exceptions (Table 2).

Table 3 summarizes the quality of each included trial based on the Jadad score. Although all the

studies were randomized, a description of the allocation sequence was only provided in 6 studies.

All the trials reported on any study withdrawals and described the completeness of follow-up.

However, none of the studies was blinded.

2.3.3 Freedom from Angina

Overall, PCI was associated with improvement in freedom from angina compared with medical

therapy (summary OR 1.69; 95% CI 1.24-2.30; p-value <0.001) (Figure 2). At the end of trial

follow-up, 73.0% of PCI patients were free from angina, compared to 63.9% of patients who

received medical therapy alone (number need to treat of 10; 95% CI 6-29;p=0.003). However,

statistically significant heterogeneity across studies was observed (p < 0.001), with an I2 statistic

of 72.7% indicating marked variation in the estimates of freedom from angina for PCI versus

medical therapy across studies.

In Figure 3a, we evaluated the impact of PCI on freedom from angina based on the studies‟

length of follow-up. PCI was associated with significant angina relief among trials of less than

one year follow-up (71.4% of PCI patients were angina free versus 64.3% of medically treated

patients) and trials of one to five years follow-up (71.3% of PCI patients were angina free versus

61.9% of medically treated patients). Among the five trials with more than five years of follow-

up, the incremental benefit of PCI versus medical therapy did not reach statistical significance.

2.3.4 Subgroup Analysis on Post Infarction Status

Our results did not change substantially with the exclusion of the six trials that enrolled

stabilized post-myocardial infarction patients (Appendix A, available at www.annals.org ). The

summary OR of 1.92 (95% CI 1.11-3.33) for these six trials, with 3010 enrolled patients, did not

Page 33: Economic Evaluation of Percutaneous Coronary Intervention in

23

differ significantly from the remaining eight trials enrolling 4808 patients (OR 1.58 [95% CI

1.07-2.33]).

2.3.5 Subgroup Analysis on Coronary Stent Use

Of the 2644 patients enrolled in trials with greater than 50% coronary stent use, the summary OR

of 1.13 (95% CI 0.76-1.68) associated with PCI versus medical therapy on angina relief was less

than the summary OR of 2.15 (95% CI 1.48-3.13) for the 5174 patients enrolled in earlier trials

with lower overall utilization of intra-coronary stents.

2.3.6 Freedom from Angina According to Enrollment Period

In Figure 3b, we stratified the trials by their final year of recruitment. Among the three trials that

recruited patients prior to 1994, 66.6% of PCI patients were angina free versus only 39.9% of

medically treated patients, for a summary OR of 3.38 (95% CI 1.89-6.04). In the six trials that

recruited patients between 1995 and 1999, 68.7% of PCI patients were free from angina

compared to 56.7% of medically treated patients (summary OR 1.72; 95% CI 1.11-2.66). In

contrast, among the patients enrolled in the five contemporary trials that recruited patients after

2000, 77.4% of PCI patients were angina free compared to 74.8% of medically treated patients

(summary OR 1.13; 95% CI 0.76-1.68).

2.3.7 Meta-Regression of Freedom from Angina and Medical Therapy

Figure 4 shows the meta-regression analysis plotting the treatment effects of PCI relative to

medical therapy versus utilization of evidence-based medications. A statistically significant

inverse relationship between freedom from angina and number of evidence-based medications

used in a trial was observed (p=0.02). In more contemporary trials where there was also greater

use of evidence-based medications, the benefit associated with PCI for symptom relief was

diminished. On average, with each additional medication class, the advantage of PCI over

medical therapy for angina relief decreased by 31% (95% CI 14%-45%). This finding remained

consistent when the threshold for defining medication utilization in a trial was varied from 30%

Page 34: Economic Evaluation of Percutaneous Coronary Intervention in

24

to 70%. In addition, we performed an additional meta-regression analysis pooling utilization

rates of medical therapy from each trial and found a similar significant inverse relationship

between medical therapy and PCI.

2.4 Discussion

Although we found that PCI when added to medical therapy was associated with an overall

improvement in angina relief compared with medical therapy alone for patients with stable

coronary artery disease, the incremental benefit of PCI on angina relief was substantially reduced

over the trial period included in our meta-analysis. The benefit in angina relief associated with

PCI was predominantly restricted to older trials, with contemporary studies showing no

significant differences between patients treated with PCI and medical therapy. One key reason

might be improvement in the proportion of medically treated patients who became angina free

over time, from 39.9% in older trials to 56.7% in intermediate trials to 74.8% in contemporary

trials. The increasing proportion of angina free patients corresponded to an increasing use of

evidence-based medical therapy among trials. Indeed, we found that there was an inverse

relationship between utilization of evidence-based therapies and efficacy of PCI in our meta-

regression analysis. These findings were robust in multiple sensitivity analyses and not overly

influenced by any one single trial, such as COURAGE. Our study suggests that improvements in

medical therapy might explain the attenuated impact of PCI on angina relief in contemporary

practice.

It is difficult to establish the exact reason for the large variation in incremental benefit associated

with PCI versus medical therapy across different time periods, as both medical and interventional

therapies have progressed substantially. Nonetheless, we were able to discount several

hypotheses. First, a potential explanation is that there may have been a greater proportion of

medically treated patients who subsequently crossed over to receive revascularization in

contemporary trials. In fact, this was not the case as we found the need to receive

revascularization was smaller in contemporary trials with only 24.6% of medically treated

Page 35: Economic Evaluation of Percutaneous Coronary Intervention in

25

patients undergoing revascularization, versus 34.6% in older trials. Second, it was unlikely to be

related to differential enrollment of asymptomatic or minimally symptomatic patients in different

time periods as the proportion of patients with Canadian Cardiovascular Society functional

classification 0 and 1 did not differ substantially between the three recruitment periods.

Our findings were also unlikely to be attributed to the introduction of coronary stents. It has

been estimated that coronary stenting is associated with approximately 50% reduction in

restenosis and future repeat coronary revascularization (118;119). Interestingly, we observed a

greater degree of angina relief in trials of balloon angioplasty as compared with trials of stenting.

This finding may suggest that advancements in medical therapy had greater impact on angina

relief in recent years than coronary stents. However, since most trials did not include patients

with drug-eluting stents, we might have underestimated the potential benefit of PCI on angina

relief in current practice where these newer devices are routinely used.

A fundamental principle of meta-analyses is that the same research question is being addressed

on a similar patient population across all included studies. Previous meta-analyses assessing the

efficacy of PCI versus medical therapy in patients with stable angina have included studies

which enrolled patients after recent myocardial infarction (14;15;95). We have previously

demonstrated that this clinical heterogeneity results in different summary estimates for mortality

(24). In our study, in which we evaluated freedom from angina as our main outcome, we also

included trials with stabilized patients with recent myocardial infarction. To evaluate the impact

of clinical heterogeneity, we performed subgroup analyses and did not find a significant

difference in the estimates for groups when stratified by myocardial infarction status. Moreover,

we repeated our meta-regression and subgroup analysis in the non-myocardial infarction group

and found similar results (Appendix A available at www.annals.org).

Our study has important implications for the contemporary management of patients with stable

coronary artery disease. Current practice guidelines recommend PCI for angina relief in patients

with stable angina without specifically alluding to the role of medical therapy (118). Our

Page 36: Economic Evaluation of Percutaneous Coronary Intervention in

26

findings suggest that in contemporary practice, many patients would respond to medical therapy

and the incremental benefit of PCI on angina relief may be substantially smaller than previously

believed. Our findings also lend support to the new appropriateness guidelines of coronary

revascularization that emphasize the need to optimize medical therapy prior to referring patients

for PCI (120).

Our analysis should be interpreted within the context of several limitations. First, we did not

have complete information on the use of medical therapy, such as prescribed dosages or long-

term adherence rates. Nevertheless, it is unlikely that adherence to medical therapy would be

substantially different between treatment groups within a randomized trial setting where all

patients were monitored and followed in a similar manner. Importantly, aside from β-blockers,

we evaluated aspirin, statin and ACE-inhibitor use, none of which are typical therapy for

relieving angina. However, as the utilization of these classes is a proxy for the aggressiveness of

overall medical therapy, it is likely that as study-level utilization of these medications improved

in more contemporary trials, the use of other anti-anginal medications also improved. Second,

we used freedom from angina as our main outcome measure, rather than angina severity.

Among patients who remained symptomatic, we cannot exclude the possibility that PCI may

have been associated a greater improvement in the severity of angina. Third, although this

represented the most comprehensive review comparing PCI to medical therapy to date on angina

relief, our conclusions may be limited by the selective reporting of freedom from angina in

clinical trials. Finally, our meta-regression analysis assessed a single covariate, based on study-

level aggregate data on medical therapy; it is possible that additional factors might confound the

relationship we found between the number of evidence based medications and freedom from

angina. For example, we did not account for potentially greater uptake of lifestyle modifications

such as increased physical activity and smoking cessation in recent trials. As such, our analysis

should be considered exploratory and hypothesis generating, not conclusive.

In summary, we found greater freedom from angina associated with medical therapy in recent

randomized trials, and a relatively limited incremental benefit of PCI for angina relief; our

Page 37: Economic Evaluation of Percutaneous Coronary Intervention in

27

results suggest this may be related to the greater utilization of evidence-based medications in

contemporary studies.

Page 38: Economic Evaluation of Percutaneous Coronary Intervention in

28

Table 1: Trial Design and Characteristics of Randomized Controlled Trials of Medical Therapy versus PCI

Trial Name Year of

recruitment Inclusion Criteria

Number

of

patients

Follow-up

duration

(years)

Mean

age

(years)

Previous

MI Stent use

CCS

angina 0

or 1

(%) (%) (%)

ACME, 1997

(27,28) 1987-1990

Stable angina, ≥ 3mm on exercise

stress test, MI within 3 months, and

stenosis > 70% in the proximal

coronary artery of one or two

vessels

328 5 60 37 0 12.5

MASS, 1995

(19,20) 1988-1991

Stable angina and stenosis > 80%

in the proximal left anterior

descending artery

144 5 56 0 0 NA

TOPS, 1992 (13) 1992

Stable patients 4 to 14 days after

MI treated with fibrinolytic

therapy; no Q wave on ECG;

negative stress test; stenosis of ≥

50% in an infarct-related coronary

artery

87 1 57 100 0 100

RITA-2, 1997 (4,9) 1992-1996

Stable patients; unstable angina

patients without symptoms for at

least 7 days, and stenosis > 50% in

at least 2 projections or stenosis >

70% in one projection in major

coronary artery

1018 7 58 47 9 46.5

AVERT, 1999 (31) 1995-1996

Asymptomatic, stable angina,

unstable angina or MI greater than

14 days; hyperlipidemic; stenosis

>50% in one or two coronary

arteries

341 1.5 59 42 39 58.7

Page 39: Economic Evaluation of Percutaneous Coronary Intervention in

29

Dakik, 1998 (12) 1995-1996

Stable patients after MI; large total

and ischemic left ventricular

perfusion defect

44 1 54 100 32 NA

SWISSI II, 2007

(11) 1991-1997

Stable, asymptomatic patients 3

months after MI; significant

ischemia on exercise test confirmed

by stress imagine; stenosis in one

to two coronary arteries

201 10.2 55 100 0 100

ALKK, 2003 (14) 1994-1997

Stable patients 8 to 42 days after

ST-segment elevation MI;

significant stenosis of an infarct-

related coronary artery

300 4.7 58 100 17 98

Bech, 2001 (5,38) 1997-1998

Stable angina; no evidence of

reversible ischemia within 2

months; stenosis >50% lesion in

coronary artery

181 5 61 25 46 12

MASS 2, 2004

(29,30) 1995-2000

Stable angina (CCS II – III) or

ischemic stress test; stenosis >70%

in multiple proximal coronary

arteries

408 5 60 46 72 22

Hambretch, 2004

(6) 1997-2001

Stable angina (CCS I – III) with

documented ischemia; stenosis ≥

75% in coronary artery

101 1 61 46 100 36

DECOPI, 2004 (15) 1998-2001

Stable patients within 15 days after

MI; total occlusion of the infarct-

related coronary artery located in

proximal segment

212 3 57 100 80 100

Page 40: Economic Evaluation of Percutaneous Coronary Intervention in

30

COURAGE, 2007

(8) 1999-2004

Stable angina; stabilized unstable

angina; stenosis > 70% in a

proximal coronary artery with

myocardial ischemia, or stenosis >

80% and classic angina

2287 4.6 62 38 94 42.2

OAT, 2006 (7) 2000-2005

Stable patients 3 to 28 days after

MI; total occlusion of an infarct-

related coronary artery; increased

risk criteria (left ventricular

ejection fraction < 50% or

proximal occlusion of a major

epicardial vessel)

2166 4 59 100 87 100

Trials abbreviations: ACME: Angioplasty Compared to Medicine; MASS: Medicine, Angioplasty, or Surgery Study; TOPS: Treatment of Post-

Thrombolytic Stenoses; RITA: Randomized Intervention Treatment of Angina; AVERT: Atorvastatin versus Revascularization Treatment ; SWISSI:

Swiss Interventional Study on Silent Ischemia; ALKK: Arbeitsgemeinschaft Leitende Kardiologische Krankenhausarzte; DECOPI: DEsobstruction

COronaire en Post-Infarctus; COURAGE: Clinical Outcomes Utilizing Revascularization and Aggressive DruG Evaluation; OAT: Occluded Artery Trial

Abbreviations: CCS, Canadian Cardiovascular Society; MI, myocardial infarction; NA, not available; PCI, percutaneous coronary intervention.

Page 41: Economic Evaluation of Percutaneous Coronary Intervention in

31

Table 2: Baseline Medication Use in the Medical Therapy and the PCI Groups*

Trial Name Aspirin (%) β-blockers (%) ACE-inhibitor (%) Statin therapy (%)†

PCI Med PCI Med PCI Med PCI Med

ACME, 1997 (27,28) 91 83 39 42 NA NA NA NA

MASS, 1995 (19,20) 75 78 52 46 30 25 42 36

TOPS, 1992 (13) 100 100 37 37 NA NA NA NA

RITA 2, 1997 (4,9) 87 87 68 65 9 11 13 12

AVERT, 1999 (31) 89 82 69 62 8 9 69 93

Dakik, 1998 (12) 100 100 84 82 NA NA 53 45

SWISSI II, 2007 (11) 98 98 37 91 35 46 33 28

ALKK, 2003 (14) 100 100 74 75 NA NA NA NA

Bech, 2001 (5,38) 92 92 62 71 NA NA 37 37

MASS 2, 2004 (29,30) 80 80 61 68 30 29 73 68

Page 42: Economic Evaluation of Percutaneous Coronary Intervention in

32

Hambretch, 2004 (6) 98 98 86 88 88 74 80 72

DECOPI, 2004 (15)‡ 83 83 81 81 58 58 82 82

COURAGE, 2007 (8) 96 95 85 89 58 60 86 89

OAT, 2006 (7) 94 97 86 89 80 80 80 82

Abbreviations: ACE, angiotension-converting enzyme. Please refer to Table 1 for other abbreviations.

* Medication use at follow-up in MASS-2 and DECOPI is shown because data on baseline medication use was not available.

† RITA-2, Dakik, OAT reported lipid lower therapy rather than statin therapy.

‡ Overall utilization rate given because no breakdown by treatment group was given in the study. Similar use of medical therapy between patients assigned

medical treatment and PCI treatment was explicitly stated.

Page 43: Economic Evaluation of Percutaneous Coronary Intervention in

33

Table 3: Study Quality and End Point Ascertainment in Randomized Controlled Trials of Medical Therapy versus PCI

Trial Name End Point Ascertainment Study Quality*

Study

Blinding

Blinding

Technique

Randomized Concealment

of Allocation

Description

of

withdrawals

ACME, 1997 (27,28) Freedom from angina no no yes no yes

MASS, 1995 (19,20) Freedom from angina no no yes no yes

TOPS, 1992 (13) Freedom from angina not reported. Mean

change in angina classification used as a

surrogate

no no yes yes yes

RITA 2, 1997 (4,9) Freedom from angina no no yes yes yes

AVERT, 1999 (31) Freedom from angina not reported.

Improvement in angina classification used as

surrogate

no no yes no yes

Dakik, 1998 (12) Freedom from angina no no yes no yes

Page 44: Economic Evaluation of Percutaneous Coronary Intervention in

34

SWISSI II, 2007 (11) Freedom from angina not reported but the

study reported angina not leading to

revascularization. Freedom from angina

estimated as the difference between number of

patients and symptomatic angina events

no no yes yes yes

ALKK, 2003 (14) Freedom from angina no no yes no yes

Bech, 2001 (5,38) Freedom from angina no no yes no yes

MASS 2, 2004 (29,30) Freedom from angina no no yes no yes

Hambretch, 2004 (6) Freedom from angina not reported.

Hospitalization and coronary angiogram for

worsening angina used as a surrogate

no no yes yes yes

DECOPI, 2004 (15) Freedom from angina no no yes no yes

COURAGE, 2007 (8) Freedom from angina no no yes yes yes

OAT, 2006 (7) Freedom from angina no no yes yes yes

* Study quality assessed based on a 5-point Jadad scale; Please refer to Table 1 for abbreviations

Page 45: Economic Evaluation of Percutaneous Coronary Intervention in

35

Figure 1. Process of study selection

310 potentially relevant citations reviewed

after electronic literature search

8 studies excluded after full

review:

4 studies did not report on

symptoms

4 studies did not distinguish

between PCI and coronary artery

bypass surgery as modalities of

coronary revascularization

14 trials included in

analysis

22 articles selected for full review

by 2 reviewers

Page 46: Economic Evaluation of Percutaneous Coronary Intervention in

36

Odds ratio and 95% CI

0.01 0.1 1 10 100

Favours

Medical Therapy

Favours PCI

Figure 2. Summary Odds Ratio of Freedom from Angina

Study name PCI Medical

Therapy

OR lower

estimate

upper

estimate

ACME 78/124 55/127 2.22 1.33 3.68

MASS 44/68 17/66 5.28 2.51 11.11

TOPS 34/42 23/45 4.07 1.55 10.69

RITA 2 130/188 105/206 2.16 1.43 3.26

AVERT 95/177 67/164 1.68 1.09 2.58

Dakik et al 18/19 21/22 0.86 0.05 14.71

SWISSI 2 85/96 73/105 3.39 1.60 7.19

ALKK 115/149 92/151 2.17 1.31 3.59

Bech et al 51/90 61/91 0.64 0.35 1.18

MASS 2 119/205 92/203 1.67 1.13 2.47

Hambrecht 43/50 50/51 0.12 0.02 1.04

DECOPI 101/109 92/103 1.51 0.58 3.92

Courage 316/423 296/406 1.10 0.81 1.49

OAT 234/263 233/257 0.83 0.47 1.47

Summary 1463/2003 1277/1997 1.69 1.24 2.30

(73.0%)

p-value:

(63.9%)

0.001

heterogeneity Q value: 47.7 df:13 I2:72.7

(p<0.001)

Page 47: Economic Evaluation of Percutaneous Coronary Intervention in

37

Page 48: Economic Evaluation of Percutaneous Coronary Intervention in

38

Page 49: Economic Evaluation of Percutaneous Coronary Intervention in

39

3 Predicting EQ-5D Utility Scores from the Seattle Angina Questionnaire in Coronary Artery Disease: A

Mapping Algorithm using a Bayesian Framework

Running Title: Mapping algorithm for EQ-5D scores from SAQ

Authors:

Harindra C. Wijeysundera1, 2,3,4

, George Tomlinson2,3,4

, Colleen M. Norris5,6

, William A. Ghali7,

Dennis T. Ko1,3,4,8

, Murray D. Krahn2,3,4,9

Affiliations: 1Division of Cardiology, Schulich Heart Centre and Department of Medicine,

Sunnybrook Health Sciences Centre, University of Toronto, Ontario, Canada; 2 Toronto Health

Economics and Technology Assessment (THETA) Collaborative, University of Toronto,

Ontario, Canada; 3Department of Medicine, University of Toronto, Ontario, Canada;

4Department of Health Policy, Management and Evaluation, University of Toronto, Ontario,

Canada;5Faculty of Nursing, University of Alberta;

6Division of Cardiology, Cardiovascular

Surgery and Public Health, University of Alberta; 7University of Calgary;

8Institute for Clinical

Evaluative Sciences, Ontario, Canada;9Faculty of Pharmacy, University of Toronto, Ontario,

Canada.

Presentation: 31st Annual Meeting of the Society for Medical Decision Making (SMDM),

October 20, 2009 at Hollywood (Los Angeles), California, USA.

Contents of this chapter have been published in Medical Decision Making: Harindra C.

Wijeysundera, et al, Predicting EQ-5D Utility Scores from the Seattle Angina Questionnaire

in Coronary Artery Disease: A Mapping Algorithm using a Bayesian Framework. Med

Decis Making. 2011 May-Jun;31(3):481-93. Epub 2010 Dec 2.

Page 50: Economic Evaluation of Percutaneous Coronary Intervention in

40

Funding:

Dr. Wijeysundera is supported by a research fellowship award from the Canadian

Institute of Health Research (CIHR). Dr. Ko is supported by a Heart and Stroke Foundation of

Ontario (HSFO) Clinician Scientist Award and a CIHR New Investigator Award. Dr. Krahn is

supported by the F. Norman Hughes Chair in Pharmacoeconomics. The funding agreement

ensured the authors' independence in designing the study, interpreting the data, writing, and

publishing the report.

Corresponding Author:

Harindra C. Wijeysundera

2075 Bayview Avenue, Suite A209D

Toronto, Ontario M4N3M5

Tel: (416)480-4527 Fax: (416)480-4657

Email: [email protected]

Word Count: 4634

Page 51: Economic Evaluation of Percutaneous Coronary Intervention in

41

28/07/2011

ABSTRACT

Background: The Seattle Angina Questionnaire (SAQ), a descriptive quality of life instrument,

is often used in coronary artery disease studies. In its current form, however, it cannot be used in

economic evaluations.

Objective: We sought to create a mapping algorithm that would allow translation of SAQ scores

into EQ-5D utility scores.

Design: Using data from the Alberta Provincial Project for Outcome Assessment in Coronary

Heart Disease (APPROACH) database, we examined the relationship between scores in each of

the five domains of the SAQ (physical limitation, anginal stability, anginal frequency, treatment

satisfaction, and disease perception) and the EQ-5D utility score. The cohort was divided into

80% derivation and 20% validation sets. Mapping algorithms were developed using simple

linear regression, and Tobit models. To account for the skewed distribution of the EQ-5D scores

and the presence of heteroscedasticity, we applied Bayesian extensions to each model by

specifying a non-constant variance for the error term. Model performance was assessed by

comparing predicted and observed mean EQ-5D scores in the validation set, in addition to the

amount of variance explained (unadjusted R2) by the model.

Results: Our cohort consisted of 1992 patients. We found that the simple linear regression

model had the best predictive performance, with an R2 of 0.38. The non-constant variance term

did not improve overall performance for any of the models. The linear regression model

accurately estimated the mean EQ-5D score in our validation set (predicted score: 0.81 versus

observed score: 0.81), as well as mean scores in subgroups stratified by symptom severity.

Page 52: Economic Evaluation of Percutaneous Coronary Intervention in

42

28/07/2011

Conclusions: We found that mean EQ-5D utility weights can be accurately estimated from the

SAQ using a simple linear regression mapping algorithm.

Page 53: Economic Evaluation of Percutaneous Coronary Intervention in

43

28/07/2011

3.1 Introduction

Formal economic evaluations play an increasingly important role in the adoption of emerging

cardiovascular drugs and technologies, given the current environment of budgetary constraint

(40;121). Current guidelines recommend that the health index of choice for cost-effectiveness

analysis is the quality-adjusted life year (QALY) (39;40;121).

Health status measurement instruments provide a numerical score representing health profile

across health domains such as physical, mental, and social well being (29-31). Such instruments

can be classified into four categories, based on whether the instrument is designed for a generic

versus a disease-specific population, and whether it is a psychometric/descriptive versus a

preference-based instrument (32). Both descriptive and preference based instruments provide a

score reflective of patients‟ assessment of their own quality of life. The preference-based

measures add a valuation component to patient report. For example the EQ-5D uses community-

based preferences as their valuation component. The valuation component or utility weight

allows for the calculation of QALY‟s. Descriptive instruments are widely used in cardiovascular

studies; the Seattle Angina Questionnaire (SAQ), in particular, is commonly employed (38;122-

127). In many circumstances, preference-based instruments are not included in cardiovascular

clinical trials or outcomes research (39). A potential limitation to the validity and credibility of

economic evaluations in cardiovascular disease is the paucity of comprehensive and current data

with which to estimate the quality of life component of quality adjusted survival in these patients

(39).

Mapping algorithms have been developed to allow the prediction of utility weights from the

scores on a descriptive instrument (32). The most common technique for mapping is called the

„transfer to utility regression method‟; in this method, a regression model is created with the

summary score from the utility-based target as the dependent variable and the individual

response data from the descriptive measure as the co-variate of interest (32). Other mapping

techniques include direct revaluation and effect size translation, both of which involve direct

utility measurement using standard gamble or time trade off (32). Finally, response mapping

techniques have been described to predict an individual patient‟s responses on each item of

Page 54: Economic Evaluation of Percutaneous Coronary Intervention in

44

28/07/2011

preference based instrument, from responses on the descriptive instrument (32). The appeal of

the transfer to utility method lies in its simplicity and minimal data requirements. This approach

has been the technique of choice in mapping algorithms, accounting for approximately 70% of

published studies (32).

Current methodological guidelines for economic evaluations recommend the use of utility

weights directly measured using the EQ-5D; in the absence of this, a validated mapping

algorithm should be used (40). However, the conversion of SAQ scores to utility weights has not

been attempted (32). Our objective was to derive a mapping algorithm using a transfer to utility

regression method, in order to predict mean EQ-5D utility weights from mean scores on the

SAQ. We believe that this will be of substantial use to health services researchers by making

data from previous studies which have measured quality of life using descriptive instruments

available for secondary use in cost effectiveness analyses.

3.2 Methods

3.2.1 Data Source:

Our study cohort consisted of consecutive patients with coronary artery disease who underwent

coronary angiography in 2004 as part of the Alberta Provincial Project for Outcome Assessment

in Coronary Heart Disease (APPROACH) database. This prospective registry collects

demographic and clinical information on all patients who undergo a diagnostic angiogram in the

province of Alberta, Canada. In addition, these patients complete the SAQ and the EQ-5D. We

restricted our cohort to 2004, when baseline surveys were routinely ascertained.

3.2.2 Measures:

3.2.2.1 Seattle Angina Questionnaire (SAQ):

The SAQ is a 19 item is a descriptive, self-administered questionnaire that focuses on symptoms

and impairments in health unique to coronary disease (38). The five dimensions of coronary

artery diseases that are measured include physical limitation, anginal stability, anginal frequency,

Page 55: Economic Evaluation of Percutaneous Coronary Intervention in

45

28/07/2011

disease perception and treatment satisfaction (38). Ordinal scores on the questionnaire are

transformed to a 0 to 100 scale, with the score of each dimension analyzed separately (38).

3.2.2.2 EQ 5D:

This scale expresses preference for health status using a single index score and it is the

recommended tool by the National Institute for Health and Clinical Excellence (NICE) for

calculation of utility weights (33;40). The EQ 5D covers five dimensions of health: mobility,

self-care, usual activities, pain/discomfort, and anxiety/depression. Each dimension is measured

using a three level ordinal scale and a unique health state is defined by a combination of scores

on each of the five dimensions (33). A utility value for each of the 243 potential health states is

calculated by applying a country-specific tariff. In this analysis, we used US tariffs, derived

using regression modeling based on direct valuation on a subset of these health states using time-

trade off methodology, with a range of possible utility values between -0.109 and 1 (33).

3.2.3 Analysis:

3.2.3.1 Missing Data

Based on previous publications, we anticipated that approximately 75-85% of patients would

have complete data on both the SAQ and the EQ-5D (72;128;129). For our primary analysis we

only report a complete-case analysis, restricted to cases without missing data (130). As a

sensitivity analysis to determine if our conclusions were overly biased by incomplete data, we

accounted for missing values, assumed to be missing at random, using Markov Chain Monte

Carlo methods to impute missing data, creating five imputed datasets (Appendix A). Our

subsequent regression modeling was performed on each data set (130). In our first multi-

imputation datasets, we only used the SAQ components and EQ-ED to estimate missing values.

In a further sensitivity analysis, we repeated our imputation including age and gender as patients

with missing data were older and more likely to be female (see Results section for details).

3.2.3.2 Model Development

We created a prediction algorithm using multi-variable regression modeling, with the utility

weight from the EQ-5D being our response variable. For our primary analysis, we designed a

self-contained mapping algorithm using only information from the SAQ, rather than including

Page 56: Economic Evaluation of Percutaneous Coronary Intervention in

46

28/07/2011

other variables such as age or gender. A self-contained algorithm has the advantage of

potentially more widespread secondary application to alternative studies/data-sets as it does not

require any additional information for a valid transformation (32). Our covariates were the

summary scores (0-100) from each of the five domains of the SAQ, which we treated as

continuous variables. Previous evidence suggests that our response variable of interest, the EQ-

5D scores is skewed towards the upper bound of 1 (32;131;132). The bounded nature of the EQ-

5D utilities also means that the variance must be smaller for higher mean scores. To assess the

importance of addressing these potential complexities, we fitted two classes of models:

i. linear regression: Yi= β0 +β1X1i + ….+ βkXki + εi (1)

ii. Tobit model: Yi*= β0 +β1X1i + …. + βkXki + εi

(2)

Linear regression (equation 1) is familiar to most researchers, and often used in the mapping

literature, despite the skewed nature of utility data. The Tobit model (equation 2) is an

econometric regression model used in the presence of censored data (132). This is of relevance

in utility studies where a substantial portion of patients have a utility score of 1. The Tobit

model assumes that despite having the same observed score at the ceiling of 1, patients with

these responses may be different, and that their true health state may vary. Yi* in equation (3)

represents a valuation of an individual‟s true health state. If a patient‟s observed EQ-5D utility is

1, the model assumes that Yi* is greater than 1 (132). Therefore, the observed utility score Yi is

given by:(132)

Yi=Yi* for Yi

*<1 and Yi=1 for Yi

*>1. (3)

As our primary objective was to create a mapping algorithm for secondary use of aggregate SAQ

scores from the published literature, we did not explore the performance of models which would

depend on original individual patient data, such as linear regression with polynomial terms, or

generalized linear regression, even though they may have theoretical advantages given the

skewed nature of the EQ-5D utilities.

Page 57: Economic Evaluation of Percutaneous Coronary Intervention in

47

28/07/2011

The linear regression and Tobit models are summarized in Table 1 (Models 1, and 4), with a

conditional distribution that is defined as normal with a constant variance (132). Both models

traditionally assume that the error term εi in (1) and (2) is normally distributed with constant

variance: εi~N(0,σ2). Previous work has shown that EQ-5D utility scores do not in fact exhibit

constant variance (132). We employed methods to extend these traditional statistical models in

order to account for non-constant variance or heteroscedasticity. First, we expressed the

variance of the error term as a function of the five domains of the SAQ (Models 2, and 5) (132).

We anticipated that variance in the utility score would decrease as scores approached the ceiling

of 1 and patients were healthier (i.e. higher SAQ scores); therefore, we expressed the variance as

inversely proportional to a linear combination of the SAQ domain scores. As an alternative

form, we expressed the variance as a linear function of age and gender (Models 3, and 6) (131).

In the linear models (1, 2, and 3), we assumed that all the observed EQ-5D utility values were

from a normal distribution, with conditional mean (μi) that was a function of each of the five

components of the SAQ, as seen in Table 1. In contrast, for the Tobit models (4,5 and 6), if the

observed EQ-5D utility was less than 1, we assumed that these observations had a normal

conditional distribution, with mean μi as defined for Models 1-3. However, if the observed EQ-

5D utility was at the upper bound of 1, we assumed that it has a conditional distribution with

mean μi greater than 1. In both cases, these conditional means were a function of the five

components of the SAQ.

3.2.3.3 Model Estimation

All model fitting was done using Bayesian methods. The posterior probability distribution for

each of the model parameters was estimated using Markov Chain Monte Carlo (MCMC)

simulation methods. For each of the models with constant variance (1, and 4), a burn-in of 5000

simulations was performed, while for the extensions with heteroscedastic error (2, 3, 5, and 6), a

longer burn-in of 10,000 simulations was required to establish convergence. Convergence of the

Gibbs sampler was determined by examining the Gelman-Rubin convergence statistic for three

MCMC sequences with different initial values. The set of initial values for each parameter of

each chain was developed using random numbers generated from a normal distribution. Non-

informative prior distributions were used for all model parameters (see Appendix B). We

Page 58: Economic Evaluation of Percutaneous Coronary Intervention in

48

28/07/2011

believed that non-informative priors were justified, given the absence of any previous literature

on mapping in our population to guide the choice of prior distributions. Given the non-

informative nature of these priors and our large sample size, we did not perform sensitivity

analyses on the choice of prior distributions.

3.2.3.4 Model Reliability and Validation

We randomly divided our cohort into an 80% derivation sample and a 20% validation sample.

To compare the performance of our six models, we calculated the proportion of variance they

explained in both the derivation and validation samples, using the unadjusted R2 statistic, where

(133;134):

R2

= 1- ∑i(yi - fi)2

/ ∑i(yi - ў)2

(4)

In equation (4), yi is the observed value for a subject, ў is the mean of the observed values and fi

is the modeled value for a subject. The modeled values are determined by applying the derived β

coefficients from each MCMC simulation iteration to the observed SAQ values. For the Tobit

models, if the predicted value, fi was > 1, a value of 1 was used in equation (4) for the calculation

of R2.

In addition, we calculated the adjusted R2, which is a modification of the R

2 that adjusts for the

number of explanatory terms in the model. It is defined as:

Adjusted R2 = 1-{(1-R

2)*[(n-1)/(n-p-1)]} (5)

In equation (5), n is the sample size, while p is the number of covariates in the model. An

additional metric of model performance was the mean of the absolute prediction error which is

defined as the absolute difference between the predicted and observed value.

As the intended purpose of our mapping algorithm is to predict the mean EQ-5D scores based on

the mean SAQ domain scores, we also compared the predicted versus observed mean EQ-5D

scores in the overall validation dataset. To assess if our mapping algorithm was able to

discriminate between patients with different symptom severities, we compared the predicted and

Page 59: Economic Evaluation of Percutaneous Coronary Intervention in

49

28/07/2011

observed mean EQ-5D scores within subgroups of increasing symptom severity based on their

SAQ scores. Subgroup 1 included the most severely symptomatic patients, with scores <25;

subgroup 2 patients were ≥25 & <50; subgroup 3 patients were ≥50 & <75; and finally, subgroup

4 included the least symptomatic patients with scores ≥75.

3.2.3.5 Secondary Analyses

As a secondary analysis, we explored if a subset of the SAQ domains would map better on to the

EQ-5D in the validation set. To identify the more appropriate SAQ domains, we examined the

correlation between the summary score in the SAQ domain and the EQ-5D score, and included

domains with moderate to large correlation coefficients (<-0.3 or >0.3). In addition, we

evaluated the impact of clinical and demographic characteristics on the predictive ability of the

mapping algorithm. We a priori selected covariates included in the APPROACH database that

would potentially affect overall health status. These included age, gender, previous malignancy,

previous myocardial infarction, diabetes, chronic obstructive pulmonary disease (COPD) and

history of renal dialysis. We restricted the secondary analyses to the best performing model

identified in the previous section.

3.2.3.6 Statistical Software

All data exploration and manipulation was done used R version 2.9.0 (2009, The R Foundation

for Statistical Computing). All model derivation and validation was done using Winbugs

Version 1.4 (Medical Research Council, UK). Multiple imputation using Markov Chain Monte

Carlo estimation was performed using PROC MI in SAS 8.1 (SAS Institute). The PROC

MIANALYZE procedure was used to combine regression coefficient estimates from each

imputed dataset, so as to allow for valid statistical inferences.

3.3 Results

3.3.1 Study Cohort

The entire 2004 APPROACH study population consisted of 2394 patients. 115 (4.8%) patients

had missing values for the EQ-5D utility weight, and 339 (14.1%) had missing values for one or

more of the SAQ components. Therefore, our complete case analysis was performed on a cohort

Page 60: Economic Evaluation of Percutaneous Coronary Intervention in

50

28/07/2011

of 1992 patients (16.8% had missing values). We randomly selected without replacement, a

group of 437 patients who represented our validation set, with the remaining 1555 patients used

for model derivation.

Baseline characteristics of the entire population, the complete case cohort and our validation and

derivation set are shown in Table 2. The mean age of the cohort was 64 years with 78% being

males. There were no substantial differences in these characteristics between the complete case

cohort, and the derivation and validation data sets. In contrast, patients with missing values for

either the EQ-5D (n=115) or SAQ components (n=339) were older, with more females.

Nonetheless, the baseline EQ-5D scores and five components of the SAQ had similar means and

interquartile ranges (IQR) across these data sets. As anticipated, the EQ-5D scores did not have

a normal distribution; they were left skewed with approximately 27% having an upper bound

value of 1. The univariate relationships between the SAQ domains and the EQ-5D scores were

explored based on their degree of correlation. The anginal frequency, disease perception and

physical limitation components of the SAQ had the strongest correlations with the EQ-5D index

with correlation coefficients of 0.47, 0.59 and 0.54 respectively. In contrast the angina stability

and treatment satisfaction domains had correlation coefficients of only 0.18 and 0.29.

3.3.2 Model Estimation

The posterior means and 95% credible intervals for the intercept and β coefficients for each of

the SAQ components are shown in Table 3 for all of the models. With improvements in angina

frequency, disease perception, physical limitation and treatment satisfaction (i.e. higher SAQ

scores), global health status also improved, as reflected by higher EQ-5D utility scores.

Somewhat surprisingly, the opposite was true for the anginal stability component; as the anginal

symptoms become more stable, utility weights appeared to decrease.

The posterior means for both the intercept and β coefficients were similar between the simple

models (1, and 4) and their respective extensions. In general, the estimates for the β coefficients

for the Tobit model were larger than the corresponding values for the linear models. This

observation is likely secondary to the Tobit model assumption that the 27% of EQ-5D

observations at the upper bound were modeled as having a true value of greater than 1. The 95%

credible intervals for the β coefficients for anginal stability and treatment satisfaction

Page 61: Economic Evaluation of Percutaneous Coronary Intervention in

51

28/07/2011

components of the SAQ included 0, whereas those for the other SAQ components were restricted

to positive values.

3.3.3 Model Reliability and Validation

The R2, adjusted R

2 and mean prediction error statistics for each model are reported in Table 4.

The linear regression models had the best performance in the validation data set. The linear

regression model (Model 1), explained 38% of the variance in the validation data set. Despite

their potential theoretical advantages, the Tobit models did not perform well (R2 of 29%).

Moreover, despite the apparent presence of non-normality and heteroscedasticity in our data, the

models that allowed for a non-constant variance did not perform better than the corresponding

simpler models.

In the validation dataset, the mean predicted EQ-5D score using Model 1 was 0.81 (central 95%

credible interval 0.80-0.82), compared to a mean observed EQ-5D score in the validation dataset

of 0.81 (standard error 0.01). In Figure 1, the predicted mean EQ-5D scores based on Model 1

was compared to the observed values within subgroups of varying symptom severity as defined

by each of the five SAQ domains. Importantly, the linear regression mapping algorithm was

able to discriminate between patients across a spectrum of different symptom severities in all

five of the SAQ domains. It accurately predicted mean EQ-5D scores in the modest and

minimally symptomatic patients, who had scores from 25 to 100 (subgroup 2, 3 and 4). In these

patients, the predicted scores were within 0.02 units of the observed EQ-5D scores. However,

the mapping algorithm tended to underestimate utility scores in the most severely symptomatic

patients (subgroup 1 with SAQ domain score < 25). This was especially true in the treatment

satisfaction domain, where the predicted mean EQ-5D score in subgroup 1 was 0.58, whereas the

observed mean score was 0.75.

In addition, we assessed the performance of the linear regression mapping algorithm in

subgroups based on age (above and below 65 years), gender and the presence/absence of a

previous myocardial infarction. In each of these subgroups, our algorithm was accurate in

predicting mean EQ-5D scores (please see Appendix C).

Page 62: Economic Evaluation of Percutaneous Coronary Intervention in

52

28/07/2011

3.3.4 Secondary Analyses

To potentially improve the predictive ability of our regression models, we developed a linear

regression model incorporating only the SAQ domains with high correlation coefficients (angina

frequency, physical limitation, disease perception). However, this model had a R2 in the

derivation set of 0.45, which was identical to that in Model 1, while in the validation dataset, the

R2

was only marginally improved to 0.39 (adjusted R2 0.38; mean absolute prediction error

0.088). The mean predicted EQ-5D score in the validation set was also identical at 0.81.

In exploring the impact of other co-morbidities and clinical characteristics, we found that age,

COPD and a history of a previous myocardial infarction were statistically significant in the linear

regression model. Nonetheless, the R2 value in the validation dataset with the inclusion of these

additional covariates improved to only 0.39 (adjusted R2 0.38, mean absolute prediction error

0.087, predicted mean EQ-5D score 0.81).

3.4 Discussion

In this study, we sought to create a mapping algorithm to predict EQ-5D utility scores from

responses on the descriptive SAQ in a cohort with coronary artery disease, comparing a number

of regression models in a Bayesian framework. We found that despite the theoretical advantages

of more complex models, the simple linear regression model had the best performance in the

validation sample. In addition, modeling a non-constant variance for the error term added only

marginal predictive benefit. Although the best-fitting model exhibited only modest predictive

ability at the individual level explaining approximately 38% of the variance in our validation

data set, it was able to accurately predict the mean EQ-5D utilities scores in patients across a

spectrum of symptom severity.

Although descriptive quality of life measures are frequently reported in cardiovascular clinical

trials, there is a lack of similarly reported utility scores. Therefore, a robust algorithm that allows

one to predict mean utility scores accurately from this large repository of descriptive quality of

life scores has the potential for wide spread utilization. In this study, we attempt to create such

an algorithm for the prediction of EQ-5D utility scores from the descriptive SAQ, using models

Page 63: Economic Evaluation of Percutaneous Coronary Intervention in

53

28/07/2011

fit in a Bayesian framework using Markov Chain Monte Carlo methods. The advantage of this

framework is that it allows us to specify the conditional distribution for the utility scores and also

specify the functional form of the variance of this conditional distribution (131;132). In the

process, we are able to account for both the skewed nature of the utility scores, using a Tobit

model, in addition to the presence of heteroscedasticity.

Despite the theoretical advantages of these other models, linear regression had the best

performance in our validation dataset. We believe that there are several explanations for this

finding. First, only 27% of the patients had EQ-5D scores at the ceiling of 1. This likely reflects

that we were studying a population with coronary artery disease. Indeed, in a healthy population,

where a substantial higher proportion may have been at the ceiling, models such as the Tobit

model, polynomial models or two-part models may have a better performance. Second, we had a

relatively large sample size at 1555. Linear regression is often robust with larger sample sizes,

and this may account for its reasonable performance despite heteroscedasticity.

Adequately performing transfer to utility mapping algorithms in the literature have R2 statistics

ranging from 0.31 to 0.66 (32). We believe that the lack of overlap between the SAQ domains

and the EQ-5D is a major contributor to the modest predictive ability of our model. The transfer

to utility regression technique of mapping descriptive quality of life instruments to utility scores

involves a key assumption, that both instruments measure the same dimensions of health status.

Thus, any difference in scores is due to the reliability of the instruments and the different weights

that each instruments assigns to these dimensions (32). This assumption is questionable when

mapping a disease-specific instrument to a generic utility instrument (32). We are attempting to

transform the score from a narrow area of the quality of life space, that of cardiac symptoms as

captured by the SAQ, to the entire health status domain as reflected by the EQ-5D. The R2 of

only 0.38 suggests that the impact of cardiac symptoms alone on global health status is limited.

An additional issue of note is that the maximum amount of variance explained by a regression

technique is constrained by the measurement error in each of the individual instruments. The

reliability of the EQ-5D on test-retest measurements is good with an intraclass correlation

coefficient of 0.73 (33). The domains of the SAQ are scored separately, with poor reliability for

anginal stability with an intraclass correlation coefficient of only 0.24 compared to good

Page 64: Economic Evaluation of Percutaneous Coronary Intervention in

54

28/07/2011

reliability for physical limitation and treatment satisfaction, with intraclass correlation

coefficients of 0.83 and 0.81 respectively (38). Therefore, assuming a perfect univariate

relationship between EQ-5D and the domains of the SAQ, we would anticipate our ideal model

to explain between 18% to 60% of the variance in the data (i.e. product of the intraclass

correlation coefficients in the EQ-5D and SAQ) (135). Although a multivariate model would

explain a greater proportion of the variance, this can never reach 100% (135). Therefore, one

explanation for modest predictive performance of our model is the inherent measurement error of

the EQ-5D and SAQ.

Given our model‟s modest predictive ability, caution should be exercised in applying this

transformation to derive utility scores for individual patients. Our objective however, was to

address the lack of utilities for secondary use in cost-effectiveness analyses. In this setting, the

mean utility score for a group of patients is of relevance, not the particular utility weights for an

individual. Thus, the anticipated use of our algorithm is to convert the mean SAQ scores from

randomized trials or registries to their corresponding mean utility scores. To this end, our

mapping algorithm was accurate. In addition, within subgroups of patient severity, as defined by

the quartiles of the SAQ domains, this mapping algorithm was able to discriminate health status.

To the best of our knowledge, this is the first study to develop a mapping algorithm for the

conversion of SAQ responses to utility weights. The only previous study that estimates utility

data in a cardiovascular population predicted utility scores from Canadian Cardiovascular

Society (CCS) angina severity score and the number of anti-anginal medications in 533 clinic

patients (39). Using a linear regression model, they found an R2 value of 0.37 (39). Unlike this

earlier model, we developed a self-contained algorithm, mapping the SAQ, a widely-used

descriptive measure of quality of life instrument with well established psychometric properties.

The advantage of this self-contained model is that its only data requirements are the mean SAQ

domain scores which are easily accessible from published sources.

Importantly, when available, utility weights directly obtained via the EQ-5D, should be used in

economic evaluations. Our mapping algorithm should be considered a second option, when such

data are not available. Our objective was to address the paucity of directly measured utility

weights in the cardiovascular literature. By evaluating the uncertainty introduced in decision

Page 65: Economic Evaluation of Percutaneous Coronary Intervention in

55

28/07/2011

models by use of our algorithm, one can potentially determine the value of further research into

directly measured utility value. We believe that this is an important focus for further work.

Our study should be interpreted in the context of several limitations. First, we only conducted a

complete case-analysis. The total proportion of missing values was 16.8%; although we did not

observe significant differences in either demographic characteristics or the scores in the EQ-5D

and SAQ between the overall cohort and the complete case dataset, it is likely that our data is not

missing completely at random (130). Instead, patients with lower quality of life are more likely

to be non-responders on a quality of life questionnaire; as such, our coefficient estimates may be

biased. Despite this, we did not find that our overall conclusions in so far as the degree of

variance explained by the linear regression model was substantially different in our imputed data

sets (Appendix A) (130).

Second, given that our intent was to create a self-contained algorithm to predict utility scores, we

did not include important covariates such as age, gender or co-morbidities in our primary

regression models. To assess the impact of ignoring these parameters, we performed sensitivity

analyses including these covariates and did not find a substantial increase in the R2. This

suggests that there is no additional information relating to global health status contained in age,

gender or co-morbidity status that is not reflected by a patient‟s subjective responses to the SAQ.

Finally, we did not explore the predictive ability of other complex regression models. Other

investigators have explored 2-part models, and censored least absolute deviation (CLAD) models

for mapping utilities (132;133;136). Although these models had greater predictive ability when

compared to linear regression in these studies, it is apparent from the literature that the ideal

model is heavily dependent on the derivation data (32;136-138). Linear regression models are

the most frequently used in mapping algorithms and despite their theoretical limitations appear to

be robust in practice (32;136-139).

In conclusion, we found that mean EQ-5D scores can be accurately predicted using a simple

linear regression mapping algorithm from the SAQ. We believe that this tool will be of

substantial use to researchers performing economic evaluations, by addressing the need for

preference-based utility weights in subgroups of patients with coronary artery disease.

Page 66: Economic Evaluation of Percutaneous Coronary Intervention in

56

28/07/2011

Disclosures:

The authors have no conflicts to disclose.

Page 67: Economic Evaluation of Percutaneous Coronary Intervention in

57

28/07/2011

Table 1: Model Specification

Model Conditional

Distribution

Specification of the Mean Specification of the Variance

1

Linear Regression

Yi~N(μi, ) μi= β0 +β1AFi + β2ASi + β3DPi + β4PLi + β5TSi

= σ2

2

Linear Regression

with specified

variance: SAQ

1/ = α0 + α 1AFi + α 2ASi + α 3DPi +

α 4PLi + α 5TSi

3

Linear Regression

with specified

variance: age

&gender

1/ = α0 + α 1agei + α 2genderi

4

Tobit

Yi*~N(μi, )

Yi=Yi* for

Yi*<1 and Yi=1

for Yi*>1

μi= β0 +β1AFi + β2ASi + β3DPi + β4PLi + β5TSi

= σ2

5

Tobit with

specified variance:

SAQ

1/ = α0 + α 1AFi + α 2ASi + α 3DPi + α

4PLi + α 5TSi

6

Tobit with

specified variance:

age & gender

1/ = α0 + α 1agei + α 2genderi

μ: mean; : variance; AF: Angina Frequency; AS: angina stability; DP: Disease Perception; PL: physical limitation; TS: treatment

satisfaction

Page 68: Economic Evaluation of Percutaneous Coronary Intervention in

58

28/07/2011

Table 2: Baseline Characteristics of Study Cohort

Entire

population

Complete-

Case Cohort

Missing Data Derivation Set Validation Set

EQ-ED SAQ

number of

patients

2394 1992 115 339 1555 437

Age (years)

(mean & IQR)

64.5

(57.0-72.7)

64.2

(56.4-72.0)

67.5

(60.1-75.03)

67.5

(59.8-74.5)

64.3

(56.7-72.0)

63.8

(55.8-72.4)

Male (%) 78.2 78.7 73.9 76.1 78.8 78.5

Proportion with

stable angina

(%)

40.9 41.6 47.0 36.3 42.6 38.0

EQ-5D

(mean & IQR)

0.81

(0.76-1.0)

0.81

(0.76-1)

NA 0.78

(0.71-1)

0.81

(0.77-1.0)

0.81

(0.71-1.0)

SAQ-AS

(mean & IQR)

71.2

(50.0-100.0)

72.2

(50.0-100.0)

68.1

(50.0-100.00)

NA 71.7

(50.0-100.0)

73.9

(50.0-100.0)

SAQ-AF

(mean & IQR)

79.9

(70.0-100.0)

78.8

(60.0-100.0)

82.2

(70.0-100.0)

NA 79.2

(60.0-100.0)

77.5

(60.0-100.0)

SAQ-DP

(mean & IQR)

62.1

(41.7-83.3)

61.5

(41.7-83.3)

61.2

(41.7-79.2)

NA 61.7

(41.7-83.3)

61.0

(41.7-83.3)

SAQ-PL

(mean & IQR)

70.3

(50.0-91.7)

70.00

(50.0-91.7)

67.9

(49.3-91.7)

NA 70.2

(50.0-91.7)

69.4

(50.0-88.9)

SAQ-TS

(mean & IQR)

87.9

(81.3-100.0)

87.5

(81.3-100.0)

88.5

(81.3-100.0)

NA 87.6

(81.3-100.0)

87.0

(81.3-100.0)

IQR: interquartile range; SAQ: Seattle Angina Questionnaire; AF: Angina Frequency; AS: angina stability; DP: Disease Perception; PL: physical limitation; TS:

treatment satisfaction

Page 69: Economic Evaluation of Percutaneous Coronary Intervention in

59

28/07/2011

Table 3: Results of Model Derivation

Model 1

(Linear Regression)

Model 2

(Linear Regression

with non-constant

variance: SAQ)

Model 3

(Linear Regression

with non-constant

variance: age &

gender)

Model 4

(Tobit)

Model 5

(Tobit with non-

constant variance:

SAQ)

Model 6

(Tobit with non-

constant variance:

age & gender)

Posterior mean

(central 95% credible interval )

Intercept 0.4388

(0.4015 to 0.4763)

0.4425

(0.4020 to 0.4824)

0.4463

(0.4095 to 0.4833)

0.3635

(0.3137 to 0.4128)

0.3512

(0.3070 to 0.3949)

0.3768

(0.3233 to 0.4186)

β AF 0.0010

(0.0007to 0.0013)

0.0011

(0.0007 to 0.0015)

0.0009

(0.0006 to 0.0012)

0.0012

(0.0008 to 0.0016)

0.0012

(0.0009 to 0.0016)

0.0011

(0.0007 to 0.0015)

β AS -0.0002

(-0.0005 to 0.0000)

-0.0002

(-0.0004 to 0.0001

-0.0002

(-0.0004 to 0.0000)

-0.0004

(-0.0006 to -0.0001)

-0.0005

(-0.0008 to -0.0002)

-0.0003

(-0.0006 to -0.0001)

β DP 0.0023

(0.0020 to 0.0027)

0.0023

(0.0020 to 0.0026)

0.0023

(0.0020 to 0.0026)

0.0032

(0.0027 to 0.0036)

0.0031

(0.0026 to 0.0035)

0.0030

(0.0026 to 0.0035)

β PL 0.0019

(0.0017 to 0.0022)

0.0019

(0.0016 to 0.0022)

0.0019

(0.0016 to 0.0022)

0.0025

(0.0021 to 0.0028)

0.0027

(0.0023 to 0.0031)

0.0023

(0.0020 to 0.0028)

β TS 0.0004

(-0.0001 to 0.0008)

0.0003

(-0.0002 to 0.0008)

0.0004

(-0.0001 to 0.0008)

0.0006

(0.0000 to 0.0011)

0.0007

(0.0001 to 0.0012)

0.0007

(0.0002 to 0.0011)

σ2 0.01412

(0.01316 to 0.01515)

0.01773

(0.0134 to 0.02099)

0.0141

(0.01408 to 0.01516)

0.02318

(0.02121 to 0.02529)

0.02476

(0.02235 to 0.02758)

0.02969

(0.02133 to 0.05094

AF: Angina Frequency; AS: angina stability; DP: Disease Perception; PL: physical limitation; TS: treatment satisfaction; σ2: variance

Page 70: Economic Evaluation of Percutaneous Coronary Intervention in

60

28/07/2011

Table 4: Model Performance

Model 1

(Linear

Regression)

Model 2

(Linear

Regression with

specified

variance: SAQ)

Model 3

(Linear

Regression with

specified

variance: age &

gender)

Model 4

(Tobit)

Model 5

(Tobit with

specified

variance:

SAQ)

Model 6

(Tobit with

specified

variance: age &

gender)

Posterior mean (central 95% credible interval)

DERIVATION DATA-SET

R2 0.45

(0.45 to 0.45)

0.45

(0.45 to 0.45)

0.45

(0.45 to 0.45)

0.41

(0.39 to 0.42)

0.40

(0.38 to 0.41)

0.42

(0.40 to 0.44)

Adjusted R2 0.45

(0.45 to 0.45)

0.45

(0.44 to 0.45)

0.45

(0.45 to 0.45)

0.41

(0.39 to 0.43)

0.40

(0.38 to 0.41)

0.42

(0.39 to 0.44)

Mean

prediction

error

0.089

(0.088 to

0.090)

0.089

(0.088 to 0.090)

0.089

(0.088 to 0.090)

0.089

(0.088 to 0.090)

0.090

(0.089 to

0.091)

0.088

(0.087 to 0.090)

VALIDATION DATA-SET

R2 0.38

(0.36 to 0.39

0.38

(0.36 to 0.39)

0.38

(0.37 to 0.39)

0.29

(0.26 to 0.32)

0.28

(0.25 to 0.30)

0.32

(0.27 to 0.35)

Adjusted R2 0.37

(0.35 to 0.38)

0.37

(0.36 to 0.38)

0.37

(0.36 to 0.38)

0.28

(0.25 to 0.31)

0.27

(0.24 to 0.30)

0.31

(0.26 to 0.34)

Mean

prediction

error

0.088

(0.087 to

0.090)

0.088

(0.087 to 0.090)

0.088

(0.087 to 0.89)

0.089

(0.087 to 0.091)

0.090

(0.088 to

0.091)

0.088

(0.087 to 0.090)

Predicted

mean EQ-

5D

0.81

(0.80 to 0.82)

0.81

(0.80 to 0.82)

0.81

(0.81 to 0.82)

0.84

(0.83 to 0.84)

0.83

(0.83 to 0.85)

0.83

(0.82 to 0.84)

R2: proportion of variance explained by the model

Page 71: Economic Evaluation of Percutaneous Coronary Intervention in

61

Figure Legends

Figure 1: For all SAQ domains subgroup1: score <25; subgroup 2: 25≤ score <50;

subgroup 3: 50≤ score <75; subgroup 4: score ≤100

Page 72: Economic Evaluation of Percutaneous Coronary Intervention in

62

62

Page 73: Economic Evaluation of Percutaneous Coronary Intervention in

63

4 Medical Therapy vs. PCI in Stable Coronary Artery Disease: A Decision Analysis and Economic Evaluation

Authors:

Harindra C. Wijeysundera1, 2,3,4

, George Tomlinson2,3,4

, Dennis T. Ko1,3,4,5

, Vladimir Dzavik3,6

,

Murray D. Krahn2,3,4,7

Affiliations: 1Division of Cardiology, Schulich Heart Centre and Department of Medicine,

Sunnybrook Health Sciences Centre, University of Toronto, Ontario, Canada; 2 Toronto Health

Economics and Technology Assessment (THETA) Collaborative, University of Toronto,

Ontario, Canada; 3Department of Medicine, University of Toronto, Ontario, Canada;

4Department of Health Policy, Management and Evaluation, University of Toronto, Ontario,

Canada; 5

Institute for Clinical Evaluative Sciences, Ontario, Canada; 6

University Health Network

– Toronto General Hospital, Ontario, Canada;7Faculty of Pharmacy, University of Toronto,

Ontario, Canada.

Page 74: Economic Evaluation of Percutaneous Coronary Intervention in

64

ABSTRACT

Context: Percutaneous coronary intervention (PCI) with either drug-eluting stents (DES) or

bare metal stents (BMS) reduces recurrent angina and decreases the need for additional

procedures compared to aggressive medical therapy alone. It remains unclear if these benefits

are sufficient to offset their increased costs and small increase in adverse events.

Objective: Economic evaluation of initial medical therapy versus PCI with either BMS or DES.

Design: Cost-utility analysis using a Markov cohort model, with lifetime time horizon, from the

perspective of the Ontario Ministry of Health and Long Term Care (third party payer).

Data Sources: We used observational data from Ontario, Canada for baseline event rates for

recurrent angina, myocardial infarction and death. Effectiveness and utility data were obtained

from the published literature, with costs obtained from the Ontario Case Costing Initiative.

Target Population: Patients with stable coronary artery disease, defined at angiography.

Interventions: Optimal medical therapy, PCI with a bare metal stent, and PCI with a drug

eluting stent.

Outcome Measures: Recurrent angina, lifetime PCI procedures, lifetime costs, quality-adjusted

life years (QALY), and the incremental cost-effectiveness ratio (ICER).

Results of Base Case Analysis: We found that medical therapy had the lowest life-time costs at

$22,952, with BMS and DES having life-time costs of $25,081 and $25,536 respectively.

Medical therapy had a quality-adjusted life expectancy of 10.10 QALY. BMS had the greatest

quality-adjusted life expectancy of 10.26 QALY, producing an ICER of $13,271 per QALY. In

Page 75: Economic Evaluation of Percutaneous Coronary Intervention in

65

contrast, the DES strategy had a quality-adjusted life expectancy of only 10.20 QALY and was

dominated by the BMS strategy.

Conclusions: In patients with stable coronary artery disease, an initial strategy of PCI with a

BMS is cost-effective. This model has potentially significant policy and clinical implications, in

so far as recommendations for initial treatment strategy in stable angina.

Page 76: Economic Evaluation of Percutaneous Coronary Intervention in

66

4.1 Introduction

In contemporary practice, the most common alternative treatment options for a patient with

symptomatic stable angina are medical therapy alone, or percutaneous coronary intervention

(PCI) with either a bare metal stent (BMS) or a drug-eluting stent (DES) in combination with

medical therapy (3).

Since its advent in 1979, PCI has been shown to be highly efficacious in the treatment of angina.

Over the last 30 years, there have been substantial advances in PCI technology, first with the

adoption of BMS, and more recently the use of DES to reduce the phenomenon of restenosis

(3;7). Recent studies have raised concerns regarding the long-term safety of DES due to a

potential increase in late myocardial infarctions (MI) from stent thrombosis due to incomplete

endothelization of stent struts (8;9). In comparison to BMS, DES have significantly higher

acquisition costs and require the prolonged administration of both aspirin and a thienopyridine to

provide dual anti-platelet coverage until adequate stent strut endothelialization has occurred

(10;11). Medical therapy has also improved dramatically over this period. The Clinical

Outcomes Utilizing Revascularization and Aggressive Drug Evaluation (COURAGE) trial

compared an initial strategy of aggressive medical therapy versus PCI with BMS in patients with

stable coronary artery disease, and found no statistically significant survival difference, although

greater symptom relief was associated with PCI (13;22). Thus, the decision to perform a PCI in

patients with stable angina is based on its effectiveness in relieving symptoms, preventing

recurrent angina and reducing repeat procedures.

From a health policy perspective, it remains unclear if the magnitude, to which PCI with either

BMS or DES, in comparison to medical therapy alone, reduces angina and subsequent

procedures, offsets the increased costs of stenting and the potential pre-procedural and long term

safety risks. The majority of previous economic evaluations have focused on the comparison of

DES versus BMS (55;68-72;140). Those studies which have included a medical therapy arm are

limited to cost-effectiveness analyses of randomized controlled trials or analyses prior to the

wide-spread adoption of stenting (48-51;82;85). The objective of this study is to evaluate the

overall clinical benefit as well as the incremental cost-effectiveness of PCI with either DES or

Page 77: Economic Evaluation of Percutaneous Coronary Intervention in

67

BMS compared to aggressive medical therapy alone for patients with stable coronary artery

disease, using a validated decision analytic model developed using observational data from

Ontario, Canada.

4.2 METHODS

4.2.1 Research Ethics Board Approval:

This study was approved by the Institutional Research Ethics Board at Sunnybrook Health

Sciences Centre, Toronto, Ontario.

4.2.2 Study Design and Outcomes:

We performed a cost utility analysis, using a Markov cohort simulation to model outcomes for a

cohort of patients with significant coronary artery disease. The cycle length of the Markov model

was 1 month. Outcomes of interest were life expectancy and quality-adjusted life years

(QALY), costs (reported in 2008 Canadian dollars) and the incremental cost-effectiveness ratio

(ICER), calculated as the incremental cost per QALY gained.

4.2.3 Economic Assumptions:

The analysis was done from the perspective of the Ontario Ministry of Health and Long Term

Care (MOHLTC), the third party payer for government insured health services in the province.

The time horizon for this analysis was life-time. All outcomes were discounted at 5% per year

(121).

4.2.4 Base Case:

The theoretical population to which our model applies is patients with stable coronary artery

disease, with symptoms of a sufficient severity to warrant coronary angiography. These patients

with angiographic confirmation of hemodynamically significant stenoses in the epicardial

coronary arteries, would be eligible for either PCI or medical therapy. The population includes

both genders with a distribution of initial angina severity, as defined by Canadian Cardiovascular

Society functional (CCS) class.

Page 78: Economic Evaluation of Percutaneous Coronary Intervention in

68

To inform the characteristics of this population, we obtained data from the Ontario Cardiac Care

Network (CCN) PCI registry, described in detail below. Importantly, we did not include patients

with unstable coronary syndromes, or patients in whom the initial anatomy was severe enough

such that coronary artery bypass grafting (CABG) would be the initial revascularization modality

of choice.

4.2.5 Data Source:

The CCN PCI registry contains detailed procedural information on patients who received stents

in the province of Ontario from December 1st, 2003 to March 31

st, 2005 (141). Long term

follow-up over 4 years in terms of rehospitalisation, repeat procedures and mortality is available

for these patients (141). This data was used to estimate baseline transition probabilities and

describe the base case cohort.

In order to adjust for baseline demographic and clinical differences between patients who

underwent PCI with BMS versus DES in the CCN registry, a propensity-score was developed to

identify well-matched pairs (141). This has been described in detail previously (141;142).

Briefly, a logistic regression model was fitted, with the dichotomous dependent variable being

receipt of a DES (141;142). The propensity score is the model-based predicted probability of

receiving a DES. A propensity-score–matched cohort was created by matching each patient in

the DES cohort with one in the BMS cohort (a 1:1 match) (141;142). A nearest-neighbour–

matching algorithm was used to match patients on the basis of the logit of their propensity score,

with matching occurring if the difference in the logits of the propensity scores was less than 0.2

times the standard deviation of the scores (the caliper width) (141;142). Using this method, 3751

pairs of well matched patients were identified. As we were only interested in patients with stable

angina, we excluded any patient with a recent MI (within 1 month).

The mean age of our cohort was 63.1 years and 70.8% were male. There was a range of initial

symptom severity, with 13.7% having only minimal symptoms (CCS class 0-1), while 21.5%

had CCS class 2, and 64.8% having severe CCS 3-4 symptoms.

Page 79: Economic Evaluation of Percutaneous Coronary Intervention in

69

4.2.6 Treatment Strategies and Model Structure:

The three strategies evaluated were optimal medical therapy alone, PCI with a BMS, or PCI with

a DES. The simplified schematic of the model structure is shown in Figure 1. Patients in the

medical therapy strategy entered the Markov process in the medical therapy sub-tree, while BMS

or DES patients entered in the PCI sub-tree. Initial angina severity for the first cycle,

irrespective of strategy, was based on the distribution of symptom severity in the CCN registry.

i. Initial Medical Therapy. The structure of the medical therapy sub-tree is detailed in Appendix

A. Potential transitions were: 1. to remain symptomatic; 2. deteriorate to more severe angina; 3.

have an MI; 4. die or; 5. transition to an angina-free health state, defined as either CCS 0 or 1

angina. A patient who was symptom free at the onset would enter the model in the CCS 0 or 1

angina free health state. The four possible transitions in the angina-free health state were: 1.

death; 2. MI; 3. recurrent angina or; 4. remaining angina-free. Recurrent angina could either be

CCS 2 or 3-4 in severity (with a probability based on the CCN patient population distribution of

initial symptom severity).

Patients in the medical therapy strategy who suffered from recurrent angina would be treated

medically. Based on the COURAGE trial, we assumed that if medical therapy was unsuccessful

in alleviating angina symptoms (i.e. transitioning the patient to CCS class 0 or 1) for 10

consecutive months, a PCI would be performed (13). This assumption was tested in our

sensitivity analysis as outlined below. Similarly, if there was deterioration in symptom severity,

a PCI would be performed immediately. The patient would then transition to the PCI sub-tree

(Appendix B), with an equal probability of a BMS or a DES stent.

ii. Initial PCI strategy with DES or BMS. All patients who underwent an initial strategy of PCI

with either BMS or DES entered the Markov process in the PCI sub-tree (Appendix B). There

was a probability of procedural failure (1%) with DES; we assumed for the base case scenario

that a BMS would be 100% deliverable (0% failure rate). If a DES could not be delivered due to

technical considerations, a BMS was used instead (3). The peri-procedural complications

included death, stroke and major bleeding. Major bleeding is relatively common and is

associated with both increased cost and poor outcomes (143). Peri-procedural stroke may result

in death, minor disability or major disability (144). The model assumed that minor disability

Page 80: Economic Evaluation of Percutaneous Coronary Intervention in

70

after a stroke was transitional and that the patient would have resolution of all neurological

defects after 1 month. After experiencing major disability from a stroke, we assumed that the

patient would not undergo any subsequent invasive cardiac procedures. If the PCI was

successful, patients entered a post-PCI angina-free (CCS 0/1) state.

In the post-PCI angina-free state, the patient could undergo four possible transitions: 1. death; 2.

MI; 3. recurrent angina or; 4. remaining angina-free. Post-PCI angina was assumed to be due to

restenosis in the original lesion, which would be treated with a DES (145). The severity of

recurrent angina was either CCS 2 or CCS 3-4, based on the distribution of initial symptom

severity in the CCN population. If a DES could not be delivered in the setting of a restenotic

lesion, balloon angioplasty alone was performed.

iii. MI. In any of the three strategies, if a patient suffered an MI, they entered a MI transition

state (Figure 1;Appendix C), with the simplifying assumption that all MIs were of sufficiently

high risk to warrant an invasive strategy with PCI (145). If the PCI was successful, the patient

entered a separate post-MI state to reflect the different quality of life and costs after a MI(146).

Post-MI patients could have a range of functional capacity, as defined by New York Heart

Association (NYHA). The initial distribution of post-MI NYHA Class 1 to 4 was derived from

observational studies (147). Patients could subsequently transition between NYHA Class, in

addition to transitioning to death, recurrent MI, or recurrent angina.

We contrasted MIs, which were all assumed to be in de novo lesions which had not been

previously stented, with very late stent thrombosis, which were by definition restricted to

previously stented segments. Very late stent thrombosis was treated by either repeat stenting with

a DES or balloon angioplasty. If a patient survived a very late stent thrombosis however, they

entered a similar post-MI state as described above.

iv. CABG: This model does not apply to patients with severe multi-vessel or left main coronary

artery disease, in whom CABG would be the initial revascularization. Instead, CABG was

restricted to patients who had failed PCI. Specifically, a patient could only undergo a total of

three stent procedures, consistent with previous studies and clinical practice (148). If a patient

had recurrent angina after their third PCI, they would enter a CABG state. A CABG could be

complicated by peri-procedural death. If the CABG was successful, the patient entered a post-

Page 81: Economic Evaluation of Percutaneous Coronary Intervention in

71

CABG angina-free state, with the same four possible transitions as the previous stable states

(Figure 1 and Appendix D). The model assumed that recurrent angina or MI after a CABG was

due to graft failure and that these events were treated with PCI (with an equal probability of

BMS or DES)

4.2.7 Probabilities and Hazard Ratios:

All baseline transition probabilities for the BMS and DES patients were obtained from the CCN

registry (141). Propensity matched pairs of patients with DES or BMS, as defined earlier were

followed prospectively to determine rates of death, MI and target vessel revascularization (141).

i. Recurrent Angina. As the CCN database does not have data on recurrent angina, target vessel

revascularization was used as a proxy for recurrent symptoms. This requires two simplifying

assumptions: first that all patients with clinically significant recurrent angina underwent a repeat

procedure, and that all patients who underwent repeat target vessel revascularization in the CCN

database did so for clinically significant recurrent angina. Data were available for 3 years of

follow-up (141). Rates of clinical restenosis peak within 6 months post stent implantation; late

restenosis is more infrequent (3;141). In order to determine the rates of recurrent angina,

conditional probabilities for recurrent angina were calculated for the BMS arm of the CCN

database at 0 to 6 months, 6 months to 1 year, 1 to 2 years and 2-3 years (Table 1). Each of these

probabilities was converted to a monthly rate for that time period, assuming an exponential

parametric survival function. As data was not available beyond 3 years, we assumed that the rate

of recurrent angina in the 2 to 3 year period would apply to the subsequent time horizon of the

model.

Rates of recurrent angina for DES were determined by first calculating hazard ratios to reflect

the effectiveness of DES compared to BMS, using data from the CCN database (141). Anti-

proliferative medication is eluted in a controlled fashion from the DES; therefore, we would

anticipate that any benefit in comparison to BMS to be greatest early after implantation and

would then reduce with time (3). Therefore, hazard ratios for DES versus BMS were calculated

for the time periods of 0 to 6 month, 6 months to 1 year, 1 year to 2 years, and then 2-3 years,

assuming a piecewise exponential survival function (Table 2). These hazard ratios were then

applied to the respective BMS recurrent angina rates for each time period. We assumed that the

Page 82: Economic Evaluation of Percutaneous Coronary Intervention in

72

hazard ratio for the period from 2 year to 3 years post-implantation applied to the subsequent

time-horizon of the model.

Medical therapy patients are not included in the CCN registry. Therefore, the hazard ratio for

recurrent angina with medical therapy in comparison to BMS was obtained from a meta-analysis

of published randomized controlled trials (13;17;21;97;98;107;109;149-153). Some publications

reported freedom from angina; we calculated the hazard ratio for recurrent angina as the inverse

of the hazard ratio for freedom from angina (153). The rate of recurrent angina for medical

therapy was calculated by applying this hazard ratio to the 2-3 year post-implantation monthly

rate of recurrent angina for BMS.

After a restenosis episode, the probability of subsequent restenosis is increased. We modeled

this increased risk by applying an additional hazard ratio of 1.1 for recurrent angina post-

restenosis (Table 2) (10). To reflect the higher risk of recurrent angina after plain balloon

angioplasty, we applied a hazard ratio of 2.85 to the baseline risk of subsequent recurrent angina

in these patients (154).

ii. Non-fatal Myocardial infarction. In a similar fashion, the conditional probability of MI for

BMS was obtained from the CCN database at time periods of 0-6 months, 6 months-1year, 1-2

years, 2-3 years and greater than 3 years (Table 1) (141). This was converted to a monthly rate

by assuming a parametric survival function. Hazard ratios for DES were calculated from the

CCN data for each of these time periods. Beyond the 3 years of CCN data, based on expert

opinion, we assumed that there was no difference between the rates of MI between DES and

BMS, except for the additional risk of late stent thrombosis with DES, as outlined below. This

assumption was tested in our sensitivity analyses. For medical therapy, we obtained an estimate

of the hazard ratio from a published meta-analysis (155). This hazard ratio of 0.81 was applied

to the baseline rate for BMS to obtain the MI rate for medical therapy.

In order to incorporate the risk of very late stent thrombosis, patients who received a DES had

the additional risk of a stent thrombosis at 0.13% per year, which persisted for 4 years after

initial stent implantation (10). The risk of death within 30 days from stent thrombosis was

estimated to be 20% (10).

Page 83: Economic Evaluation of Percutaneous Coronary Intervention in

73

iii. Death. Baseline all-cause death monthly rates were based on observed mortality rates from

the CCN database based on cumulative 4 year mortality for patients who received a BMS or DES

stents (Table 1) (141). In Table 1, the conditional probability for death at of 0-6 months, 6

months-1year, 1-2 years, 2-3 years and 3-4 years is shown for both BMS and DES strategies.

For the medical therapy patients, we applied a hazard ratio to the baseline BMS mortality rate at

each time interval, based on a meta-analysis of the published literature (155).

Beyond the first 4 years of observed CCN data, we assumed that mortality rates for the BMS

group would return to age-gender specific mortality rates for the Ontario population, which were

obtained from life-tables. A composite Ontario specific mortality rate was estimated based on the

mean age of the CCN population and the proportion that was male versus female. For DES

patients, beyond the first 4 years of observed CCN data, we applied a hazard ratio from the

literature to the baseline BMS mortality rate (Table 2) (155). To incorporate both the increased

early mortality post-MI, and the fact that post-MI mortality would return to baseline with time,

we applied a post-MI mortality hazard ratio that ranged from 1.71 to 1, decreasing as time

elapsed from the incident event (Table 2) (156).

4.2.8 Utilities:

Health state utility weights were obtained from the published literature (Table 3). We attempted

to choose a set of utilities with consistent scaling methods. Where possible, we incorporated

utility weights that were measured using a time trade off methodology. We assumed that the

utility weights for patients who were treated with any of the three strategies who achieved

similar symptom relief, as quantified by the CCS class, would be identical. The utility weight

associated with CCS 0-1 symptoms (i.e. the angina free state) was 0.94, while that for CCS 2,

and CCS 3-4 was 0.832 and 0.533 respectively. During the 1 month cycle in which a patient

underwent a PCI, a utility decrement of 0.06 was applied (10).

If a patient suffered an MI, a utility decrement of 0.13 was applied for that cycle. Post MI

utilities were based on NYHA functional class after discharge from hospital. These ranged from

0.91 for patients with no heart failure symptoms (NYHA 0-1), to 0.8 for NYHA class 2, and 0.3

for NYHA class 3/4.

Page 84: Economic Evaluation of Percutaneous Coronary Intervention in

74

4.2.9 Costs:

Costs for each health state included direct medical costs associated with medication,

hospitalization, and physician services. We only included medication costs for patients above

the age of 65 years, as universal drug coverage by the MOHTLC is only provided for this age

group. The two broad categories of medication costs were those associated with the secondary

treatment of coronary artery disease, and that for heart failure. We assumed the proportion of

patients on aspirin, statins, B-blockers, ACE/ARB, nitrates and calcium channel blockers was

identical to that in the PCI or medical therapy arms of the COURAGE trial (see table 4)

(3;13;145). After implantation of a BMS, 1 month of clopidogrel was required(145). In

contrast, implantation of a DES required 12 months of clopidogrel(145). Patients who survived a

MI would also require 12 months of clopidogrel after hospital discharge(145). Patients with

symptomatic heart failure (NYHA 2-4) would require HF medications, with 25% on a diuretic

and 25% on aldactone. All unit costs for medications were obtained from the Ontario Drug

Benefit Formulary (2008), and we included a dispensing fee of $12 for a 3 month supply. For

simplicity, we assumed uniform unit costs for each class of medications based on the following:

all patients who were on aspirin were on 81mg once a day, those on statin were on atorvastatin

80 mg per day, those on ACE were on ramipril 10 mg per day, those on B-blockers were on

metoprolol 200 mg per day, those on calcium channel blockers were on amlodipine 10 mg per

day and those requiring topical nitrates were on 0.4 mg/hr of nitrodur. For heart failure patients,

we assumed a furosemide dose of 40 mg per day and an aldactone dose of 25 mg per day.

Stent costs were assumed to $450 for a BMS and $1725 for a DES, as per the manufacturer.

Based on the CCN registry, we assumed that 1.46 stents were placed on average for each PCI

procedure. For clinical events such as a PCI, CABG or a hospitalization for a non-fatal MI, unit

costs were obtained from the Ontario Case Costing Initiative (OCCI). This is a province wide

initiative to develop case costs for clinical events. We used International Classification

Diagnostic Code 10 (ICD) to identify typical case costs associated with each health state in the

model and the average length of stay in 2008. The mean costs and association standard deviation

(SD) are summarized on Table 4.

The OCCI does not include physician services. We estimated that physicians would visit

inpatients every day of their hospitalization, while outpatients would see a physician every

Page 85: Economic Evaluation of Percutaneous Coronary Intervention in

75

month if symptomatic. We assumed that asymptomatic patients would not visit a physician. All

unit cost for physician services were obtained from the Ontario Physician Schedule of Benefit

2008 version. Each of these is specified in Table 4.

4.2.10 Verification and External Validation:

As outlined, our prognostic and therapeutic estimates for long-term outcomes in the baseline

analysis were derived from the Ontario CCN PCI registry (141). In order to verify the model

structure and transition probabilities, we compared the model-predicted cumulative probability

of mortality and MI in the BMS and DES strategy at serial time points to the respective values in

the CCN database. The cumulative probabilities for mortality and MI were available up to 4.3

years in the CCN registry. As seen in table 5, the model predicted values were all within the

95% confidence intervals of the CCN mortality and MI data.In a similar fashion, we compared

the model predicted probabilities of recurrent angina and MI to that in the CCN database for both

the BMS and DES strategies. There was excellent agreement between model predicted and

actual values at serial time points (table 5).

For external validation of the medical therapy strategy, we compared predicted recurrent angina

rates from the model at 4.6 years to that seen in the COURAGE trial. 25.4% of patients in the

COURAGE trial underwent revascularization with PCI at 4.6 years for symptomatic angina; our

model predicted 24.0% of medically treated patients undergoing PCI at the same time-point (13).

For comparison, 14.3% of PCI patients in the COURAGE trial, of whom 97% received an initial

BMS, underwent repeat PCI at 4.6 years, with our model predicting 15.7% (13).

4.2.11 Sensitivity Analysis:

i. Subgroup analyses. We anticipated that the efficacy of both DES and BMS in reducing

recurrent angina would vary based on patient and lesion characteristics. Previous studies have

shown that the risk of restenosis is greatest in diabetic patients with long lesions (greater than 20

mm) and small arteries (< 3 mm). In order to account for this heterogeneity in our patient

population, we performed pre-specified subgroup analyses in patients with all combinations of

diabetes, small/large arteries and long/short lesions. Each of these 8 categories was validated

against the CCN data, based on repeat revascularization rates over 3 years, and separate ICERs

were calculated (Appendix E).

Page 86: Economic Evaluation of Percutaneous Coronary Intervention in

76

ii. Deterministic. 1-way sensitivity analyses were performed on all model parameters. The

ranges for the sensitivity analysis were obtained using the 95% confidence intervals from the

source documentation (Table 1, 2, 3, 4). A willingness to pay threshold of $50,000 per QALY

was used for all sensitivity analyses(157).

iii. Scenario: In order to evaluate the impact of the initial symptom severity, we conducted three

scenario analyses, where all patients were CCS 0-1, CCS 2, or CCS 3-4 on entering the model.

iv. Probabilistic. Parameter uncertainty was ascertained using a probabilistic sensitivity analysis,

using second-order Mont-Carlo simulation. Beta distributions were fitted for all probabilities and

utilities based on mean and standard deviations from source documentation. Gamma

distributions were fitted to all costs. If standard deviations were not available for costs, we used

a standard deviation that was half of the mean. Normal distributions were used to model natural

logarithms of all hazard ratios. We performed 1000 trials, with values drawn independently at

random from all 76 distributions, in order to produce a cost-effectiveness acceptability curve.

All modeling was performing using the TreeAge Pro Suite 2007 Software Release 1.5 (TreeAge

Software Inc. Williamstown MA). Propensity-matched analyses were performed using SAS

Version 9.1 (SAS Institute Inc, Cary, NC). All meta-analyses were performed using

Comprehensive Meta-Analysis Software Version 2.2 (Biostat, Inglewood NJ).

4.3 Results

4.3.1 Base-case:

Over the time horizon of the model, we estimated that approximately 19.5% of patients who

underwent initial medical therapy would suffer an MI, in comparison to 22.3% or 22.9% of

patients in the BMS and DES strategies respectively (Table 6). For recurrent angina, we found

24.0% of medical therapy patients would undergo revascularization with a PCI for inadequately

controlled symptoms. In contrast, 15.8% of patients who underwent initial PCI with a BMS, and

12.3% of patients who underwent initial PCI with a DES would require a subsequent procedure

for restenosis (Table 6).

Page 87: Economic Evaluation of Percutaneous Coronary Intervention in

77

The results of our base-case analysis are found in Table 7. We found that patients in the medical

therapy strategy had the lowest discounted life-time costs at $22,952, with those in the BMS and

DES strategies having life-time costs of $25,081 and $25,536 respectively. Despite the medical

therapy group undergoing more subsequent procedures, the greater life time cost for the PCI

strategies was likely driven by the cost associated with the initial PCI procedure and the cost

associated with the 2.8%-3.4% increase in myocardial infarctions.

Medical therapy patients had a life-expectancy of 17.47 years with a discounted quality-adjusted

life expectancy of 10.10 QALY. BMS had the greatest discounted quality-adjusted life

expectancy of 10.26 QALY. The difference in the life expectancy between the two groups was

0.13 years, while the difference in QALY‟s was 0.16. This suggests that the difference was due

to both greater symptom control with BMS and an improvement in overall long term survival.

An initial BMS strategy had an ICER of $13,271 per QALY gained compared to medical

therapy.

In contrast, the DES strategy had a quality-adjusted life expectancy of only 10.20 QALY; as

such, it was dominated by the BMS strategy, given that DES was more expensive but less

effective than BMS. The difference in life-expectancy between the BMS and DES groups was

identical to that for QALY‟s, suggesting this was principally a difference in long term survival.

4.3.2 Sensitivity Analysis:

i. Subgroup Analyses. Results of the analyses based on combinations of diabetes status, lesion

length and vessel size are found in Table 8. In non-diabetic patients, regardless of lesion/vessel

characteristics, BMS remained cost-effective, while a DES strategy was dominated. In contrast,

in all of the diabetic subgroups, the BMS strategy had greater life-time costs compared to DES.

Although DES has a favourable ICER compared to medical therapy in the diabetic subgroups,

BMS remained the most economically attractive option, except in diabetic patients with long

lesions and small vessels. In this sub-group, which represented patients at the highest risk for

restenosis and recurrent symptoms, the DES strategy was cost-effective compared to medical

therapy, with an ICER of $18,286 per QALY gained. In contrast, BMS had an ICER of $61,985

per QALY gained compared to DES.

Page 88: Economic Evaluation of Percutaneous Coronary Intervention in

78

ii. Deterministic. 1-way sensitivity analyses were performed on all probabilities (Table 1),

hazard ratios (Table 2), utilities (Table 3) and costs (table 4). The model was robust to the range

shown for the majority of these variables. Varying the initial age (52-75 years), discounting rate

(0-6%), the duration for risk of very late stent thrombosis (18 months-10 years), or the maximum

duration of medical therapy despite ongoing symptoms (3-24 months) did not alter our base case

conclusions.

Based on our previous work, when restricted to contemporary trials, there was only slight

difference between medical therapy and PCI in the occurrence of recurrent angina, with a hazard

ratio of 1.13 (153). When evaluated across this range in 1-way sensitivity analysis (Table 2), our

conclusions remained consistent.

However, the model was sensitive to 2 important parameters: a) the hazard ratio for mortality of

the medical therapy compared to BMS, and; b) the hazard ratio for mortality of DES compared

to BMS.

Compared to the base case value of 1.11, if the hazard ratio for mortality for medical therapy

versus BMS was less than 1.05, medical therapy was the dominant strategy over either of the PCI

options. If the hazard ratio of DES versus BMS for mortality was less than 1.02 (base-case

1.06), DES was the cost-effective strategy.

iii. Scenario. The results of the scenario analyses based on initial symptoms severity is shown in

Table 9. In the scenario where all patients are minimally symptomatic (CCS 0-1), the ICER for

BMS compared to medical therapy is substantially greater than the base case, at $43,739 per

QALY gained. In the scenario of greater initial symptom severity with all patients being either

moderately (CCS 2) or severely symptomatic (CCS 3-4), the BMS strategy becomes more cost-

effective, with ICER‟s of $10,584 and $8,914 respectively. In these scenarios, the cost and

effectiveness of the PCI strategies do not change compared to the CCS 0-1 scenario; the

differences are solely due to a decrease in effectiveness and increase in cost associated with the

medical therapy arm.

iv. Probabilistic. Our probabilistic sensitivity analysis showed that there was a substantial

amount of parameter uncertainty as seen in Figure 2. At a willingness to pay threshold of

Page 89: Economic Evaluation of Percutaneous Coronary Intervention in

79

$50,000 /QALY, BMS was the preferred strategy in approximately 40% of the simulations. The

acceptability curve for BMS reached an asymptote of approximately 45% at higher willingness

to pay thresholds (Figure 2). At a willingness to pay threshold of $50,000 per QALY, either of

the PCI strategies (BMS or DES) was the optimal strategy in 70% of the simulations; this

reached an asymptote of 81% at higher thresholds. In contrast, at willingness to pay thresholds

lower than $20,000/QALY, medical therapy was the preferred strategy in the majority of the

simulations.

4.4 Discussion

In this economic evaluation of medical therapy versus PCI for coronary artery disease, we

developed a comprehensive, well-validated state transition model based on Ontario

epidemiologic and cost data. We found that an initial strategy of PCI with a BMS stent was cost

effective with an ICER of $13,271 /QALY gained. This finding was consistent across the

spectrum of clinically relevant subgroups, as defined by diabetic status and lesion characteristics.

DES was only cost-effective in the highest risk group, those with diabetes with long lesions and

small arteries. Although these results were robust to a wide range of deterministic sensitivity

analyses, our conclusions must be interpreted in the context of substantial parameter uncertainty.

We found that although PCI was in general the preferred option compared to medical therapy,

the choice of DES versus BMS was less certain. This highlights the need for further research to

both understand the nature of this uncertainty and further refine parameter estimates.

Nonetheless, the results of our analysis, based on the best current evidence, should aid both

clinicians and policy makers in the initial choice of therapy for patients with stable angina.

Due to a number of recent and highly controversial studies, there is significant debate as to the

appropriate initial strategy in patients with symptomatic stable angina and a documented

significant coronary artery lesion (8;9;141;158). This controversy has arisen because, in

comparison to medical therapy, PCI reduces angina but does not have a statistically significant

impact on mortality (13). This symptomatic improvement is limited by the phenomenon of

restenosis (3;13;145). Although DES are more efficacious than BMS in angina relief, there is

concern about long term safety and cost (158). Multiple previous economic evaluations have

Page 90: Economic Evaluation of Percutaneous Coronary Intervention in

80

concentrated on comparing BMS to DES, without incorporating a medical therapy arm

(54;62;71;72;159-162). In light of recent evidence that medical therapy may be equivalent to

PCI for many patient populations, including a medical therapy arm is both relevant and

important.

Our model found that PCI with a BMS is cost effective over medical therapy. Our a priori

hypothesis was that this decision would be driven by the magnitude to which subsequent

revascularization procedures would be reduced by PCI. Surprisingly, our model was critically

sensitive to the relative mortality estimates for both medical therapy and DES in comparison to

BMS, and not very sensitive to the frequency of revascularization. Despite the large number of

studies that have evaluated this area, there remains a substantial uncertainty in these point

estimates as reflected by their wide confidence intervals (13;155;163-165).

Of the 4 previous full economic analyses comparing medical therapy to invasive therapy with

stenting, 3 were randomized trials (48-51). The Trial of Invasive versus Medical Therapy in the

Elderly (TIME) suggested that PCI was cost-effective over a time horizon of 1 year, while the

analysis from the COURAGE trial showed that initial PCI with BMS was not, with an ICER of

$168,000 per QALY gained (48;51). The third trial, an economic analysis of the Bypass

Angioplasty Revascularization 2 Diabetes (BARI2D) study, included only diabetic patients, and

again concluded that PCI (24% DES use) was not cost-effective over a 4 year period (50). The

final study was a retrospective observational analysis, comparing the cost and consequences over

6 years in 385 patients deemed appropriate for PCI, categorized based on initial medical versus

PCI versus surgical treatment strategy. This analysis found that PCI was not cost-effective. We

believe our contrasting conclusions have a number of potential explanations.

First, while the majority of previous studies had a short time-horizon, our model had a life-time

time horizon (54;62;71;159-162). The conclusions of any model comparing medical therapy to

either DES or BMS are dependent on the degree to which repeat procedures are avoided.

Multiple randomized controlled trials with long term follow-up have suggested a sustained

benefit of DES in preventing recurrent angina over at least 4 years; therefore, a model with a

shorter time-horizon may potentially produce biased conclusions about long-term cost-

Page 91: Economic Evaluation of Percutaneous Coronary Intervention in

81

effectiveness (158). Importantly, a longer time horizon will capture relevant mortality effects.

Variation in these parameters had the greatest impact on our conclusions.

Second, as the benefit of PCI is to reduce subsequent angina, the initial symptom severity of the

patient population is key, with greater benefit anticipated for patients who are more

symptomatic. The critical importance of the initial symptom severity is reinforced in the

scenario analysis shown in Table 9, where greater initial symptom severity was associated with

improved cost-effectiveness for the PCI strategy. Our base-case cohort was very symptomatic at

onset, with 64% having severe CCS 3-4 symptoms, similar to the proportion in the TIME study

where 76% had severe symptoms (48). This reflects the fact that our target population consisted

of patients with stable angina, who warranted coronary angiography. In contrast, in the

COURAGE trial, only 21% had severe symptoms; in the BARI2D trial, only 8.6% had such

severe symptoms (13;50;51;166).

Third, ours is the only decision analytic model validated against a large observational dataset.

Although randomized clinical trials provide unbiased estimates of efficacy, they are limited by

highly restrictive enrolment criteria. For example, in the COURAGE trial, only 6.4% of

screened patients with coronary disease were enrolled in the trial (13). As such, both costs and

consequences from such studies are not generalizable to real world practice.

Nonetheless, our model must be interpreted within the context of several important limitations.

Our conclusions are limited to patients who have undergone initial angiography to define

coronary anatomy. This likely reflects a population with more severe symptoms and disease, and

as such, our conclusions must not be extended to patients who would not warrant invasive

investigation. A key strength of our analysis is that our transition probabilities are derived from

real world data. However, such data has the inherent potential for confounding. Although we

have used propensity matching in order to mitigate this risk, we cannot rule-out persistent bias.

Moreover, the real world data utilized was restricted to BMS and DES patients; we used

estimates from the literature to model the outcomes of medical therapy patients.

Finally, as is reflected by the need to use multiple data sources, we found a substantial amount of

parameter uncertainty as reflected in our acceptability curves. Our analysis suggests that based

on the best evidence currently available, BMS is the cost-effective option. However, as seen in

Page 92: Economic Evaluation of Percutaneous Coronary Intervention in

82

our probabilistic sensitivity analysis, a non-trivial proportion of simulations had either DES or

medical therapy as the preferred option. This highlights the need for further analyses to further

delineate this parameter uncertainty, using expected value of perfect information (EVPI) and

expected value of partial perfect information (EVPPI) methods. In this manner, one can provide

direction as to where further research efforts should be focused to reduce this uncertainty.

In conclusion, this economic evaluation of medical therapy versus PCI in symptomatic coronary

artery disease found that an initial strategy of PCI with BMS was cost-effective. Through the

use of a Markov process, this study puts into context the recent controversies regarding the most

appropriate initial treatment for patients with stable angina and may have significant policy and

clinical implications.

Page 93: Economic Evaluation of Percutaneous Coronary Intervention in

83

Acknowledgement:

Dr. Wijeysundera is supported by a research fellowship award from the Canadian

Institute of Health Research (CIHR). Dr. Ko is supported by a Heart and Stroke Foundation of

Ontario (HSFO) Clinician Scientist Award and a CHIR New Investigator Award. Dr. Krahn is

supported by the F. Norman Hughes Chair in Pharmacoeconomics. The funding agreement

ensured the authors' independence in designing the study, interpreting the data, writing, and

publishing the report. The authors acknowledge that the clinical registry data used in this

publication are from the Cardiac Care Network of Ontario and its member hospitals. The Cardiac

Care Network of Ontario serves as an advisory body to the MOHLTC and is dedicated to

improving the quality, efficiency, access and equity of adult cardiovascular services in Ontario,

Canada. The Cardiac Care Network of Ontario is funded by the MOHLTC. This study was

supported by the Institute for Clinical Evaluative Sciences (ICES), and the Toronto Health

Economics and Technology Assessment (THETA) collaborative, which are funded by an annual

grant from the Ontario Ministry of Health and Long-Term Care (MOHLTC). The opinions,

results and conclusions reported in this paper are those of the authors and are independent from

the funding sources. No endorsement by ICES or the Ontario MOHLTC is intended or should be

inferred.

Page 94: Economic Evaluation of Percutaneous Coronary Intervention in

84

Table 1: Probabilities

* Conditional probabilities, defined as probability of event occurring in time-interval, conditional on no event

by the end of the previous time-interval

** Pooled estimate from observational studies and controlled trials.

† Threshold value above which BMS is no longer cost-effective is shown in brackets

Variable Reference Value low high sensitive

ANGINA *Recurrent Angina after BMS implantation:

CCN (141)

0-6 months 0.0769 0.0657 0.0898 No

0.5-1 year 0.0309 0.0286 0.0333 No

1-2years 0.0172 0.0161 0.0185 No

2-3 years 0.0172 0.0152 0.0193 No

Angina-relief with medical therapy Randomized Controlled Trial (109) 0.410 0.334 0.485 No

Deterioration of symptoms with medical therapy Randomized Controlled Trial(109) 0.198 0.116 0.28 No

MI

*MI after BMS implantation:

CCN (141)

0-6 months 0.0231 0.0178 0.0301 no

0.5-1 year 0.0072 0.0062 0.0080 no

1-2years 0.0103 0.0091 0.0114 no

2-3 years 0.0106 0.0096 0.0117 no

NYHA 0/1 post MI Randomized Controlled Trial (147)

0.733 0.724 0.741 no

NYHA 2 post MI 0.211 0.203 0.218 no

Improvement in NYHA class post-MI Randomized Controlled Trial (167) 0.18 0.13 0.22 no

very late stent thrombosis with DES Pooled estimate**

(10) 0.0013 0 0.004 no

death after very late stent thrombosis Pooled estimate**

(10) 0.2 0.12 0.5 no

DEATH

*Death after BMS implantation

CCN

0-6 month 0.0204 0.0154 0.027 no

0.5-1 year 0.0135 0.0118 0.0151 no

1-2 year 0.0225 0.0203 0.0246 no 2-3 year 0.0211 0.0195 0.0229 no 3-4 year 0.0187 0.0167 0.0211 no

Death after DES implantation

CCN

0-6 month 0.0156 0.0113 0.0215 no 0.5-1 year 0.0055 0.0048 0.0064 no 1-2 year 0.0163 0.0144 0.0183 no 2-3 year 0.0166 0.0152 0.0182 no 3-4 year 0.0258 0.0224 0.0296 no

PROCEDURAL

PCI failure with DES Randomized Controlled Trial: (11) 0.01 0 0.05 no

Peri-procedural PCI death Registry(168) 0.0049 0.0047 0.0051 no

Peri-procedural PCI bleed Registry(168) 0.021 0.0205 0. 0214 no

Peri-procedural CABG death Consensus guidelines(169) 0.01 0.002 0.053 no

Peri-procedural PCI stroke Registry(168;170) 0.0018 0.0016

6

0.0019

4 no

In-hospital death after PCI stroke Registry (170) 0.221 0.10 0.50 no

Major disability post-stroke Registry (144) 0.08 0.05 0.12 no

Page 95: Economic Evaluation of Percutaneous Coronary Intervention in

85

Table 2: Hazard Ratios

Hazard Ratio of Reference Value Low High Sensitive

ANGINA

Recurrent angina for DES versus BMS:

CCN(141)

6 months 0.459 0.428 0.494

no

1 year 0.710 0.690 0.730 no

2 year 1.10 0.44 1.2 no

3 years 1.00 0.44 1.14 no

Recurrent angina for medical therapy versus BMS

(3 years)

Meta-analysis

(13;17;21;97;98;107;10

9;149-153)

1.69 1.13 2.30 no

Recurrent angina for CABG versus BMS (3 years) Meta-analysis (171) 0.27 0.14 0.51 no

Recurrent angina after prior episode of restenosis Randomized

Controlled trials (10) 1.1 1 1.6 no

Recurrent angina after balloon angioplasty Meta-analysis (154) 2.85 2.0 4.0 no

MI

MI for DES versus BMS:

CCN (141)

6 months 0.694 0.655 0.731 no

1 year 1.036 1.026 1.056 no

2 year 1.517 1.510 1.526 no

3 years 1.839 1.814 1.842 no

> 3 years Assumption 1 0.68 1.26 no

MI for Medical Therapy versus BMS (3 years) Meta-analysis (155) 0.81 0.57 1.13 no

MI for CABG versus BMS (3 years) Meta-analysis (171) 1.1 0.65 1.88 no

Death

Long Term Death for DES versus BMS Meta-analysis (155) 1.06 0.71 1.58 Yes (1.02)

Death for medical therapy versus BMS Meta-analysis (155) 1.11 0.86 1.42 Yes (1.05)

*

Death post-MI:

0-5 years

5-10 years

Prospective Cohort

study (156)

1.7

1.2

1

1

2.4

1.8

no

Death for CABG versus BMS Meta-analysis (166) 0.91 0.82 1.02 no

Death for stroke with major disability Registry (172) 9.8 4.6 21 no

* Threshold value below which BMS strategy is no longer cost-effective with a WTP threshold of $50,000

per QALY

Page 96: Economic Evaluation of Percutaneous Coronary Intervention in

86

Table 3: Utilities

Health State Value Low High Sensitive

Angina-free CCS 0/1 (173) 0.94 0.72 0.99 no

Angina CCS 2 (173) 0.832 0.612 0.99 no

Angina CC3 (173) 0.533 0.31 0.75 no

Post-MI NYHA 1 (174) 0.91 0.88 0.95 no

Post-MI NYHA 2 (174) 0.8 0.51 0.99 no

Post-MI NYHA 3(174) 0.3 0.1 0.59 no

Major stroke (144) 0.5 0.3 0.7 no

*Utility decrement of PCI (10) 0.06 0 0.1 no

*Utility decrement of PCI

bleed (146) 0.04 0.03 0.1 no

*Utility decrement of MI

(10;173) 0.13 0.05 0.42 no

*Utility decrement of Minor

stroke (144) 0.2 0.1 0.3 no

* utility decrement is short term 1-time decrement over 1 cycle (ie 1 month)

Page 97: Economic Evaluation of Percutaneous Coronary Intervention in

87

Table 4: Costs

Parameter Description Reference Cost Low High Sensitive

Medical therapy

95% on Aspirin, 62% on ACE, 89% on statin, 89%

on B-blocker, 43% on calcium channel blocker,

72% on topical nitrates

Medication costs from

Ontario Drug Benefits

(ODB) Formulary 2008

$123 62 186 no

Post-PCI Medication

96% on Aspirin, 60% on ACE, 86% on statin, 85%

on B-blocker, 40% on calcium channel blocker,

62% on topical nitrates

$117 58 174 no

Heart Failure

Medication

Similar as for post-PCI patients; if NYHA 2-4,

then 25% on furosemide, and 25% on aldactone $0.94 0. 2 no

Clopidogrel $74 0 100 no

DES stent Unit cost * 1.46 stents per case Manufacturer

$2519 500 3800 no

BMS stent $657 100 1000 no

Angina: Ambulatory cost + OHIP physician billing

Ontario Case Costing

Initiative (OCCI)

for 2008

and Physician Schedule

of Benefits, 2008

$707 26 1847 no

Transfusion Case cost for transfusion $176 33 1125 no

MI Case cost for hospitalization + OHIP billing $6637 155 18409 no

PCI Case cost for PCI hospitalization + OHIP billing

$8983 235 16578 no

CABG Case cost for CABG hospitalization + OHIP

billing $24,483 6828 45400 no

Angiogram:

Hospital funding for diagnostic angiogram + OHIP

physician billing $1433 0 6000 no

Stroke (in-hospital) Case cost for stroke hospitalization + OHIP

physician billings $8573 319 20735 no

Major Stroke

(ambulatory)

$592 61 1000 no

Page 98: Economic Evaluation of Percutaneous Coronary Intervention in

88

Table 5: Validation

6 months 1 year 2 year 3 year 4.3 years

Mortality

(%)

BMS

Model 2.03 3.34 5.50 7.51 10.11

CCN

Database

2.04

(1.54-2.70)

3.36

(2.70-4.17)

5.53

(4.67-6.53)

7.53

(6.53-8.67)

10.15

(8.63-11.93)

DES

Model 1.56 2.18 3.72 5.29 10.11

CCN

Database

1.56

(1.13-2.15)

2.11

(1.60-2.78)

3.71

(3.02-4.56)

5.32

(4.49-6.30)

10.33

(7.96-13.34)

Recurrent

Angina

(%)

BMS

Model 6.34 9.42 11.38 13.14 15.44

CCN

Database

6.38

(5.45-7.45)

9.43

(8.30-10.69)

11.38

(10.15-12.75)

12.90

(11.52-14.43)

NA

DES

Model 3.02 5.30 7.38 9.37 11.92

CCN

Database

2.98

(2.37-3.75)

5.17

(4.35-6.15)

7.60

(6.60-8.76)

9.40

(8.20-10.76)

NA

MI

(%)

BMS

Model 2.29 3.06 4.17 5.28 6.36

CCN

Database

2.31

(1.78-3.01)

3.01

(2.39-3.79)

4.12

(3.38-5.01)

5.25

(4.41-6.26)

5.98

(5.05-7.08)

DES

Model 1.73 2.50 4.14 6.06 7.24

CCN

Database

1.61

(1.17-2.21)

2.34

(1.80-3.04)

4.03

(3.30-4.91)

6.10

(5.19-7.15)

6.98

(5.88-8.29)

NA: not applicable; 95% Confidence intervals in brackets.

Page 99: Economic Evaluation of Percutaneous Coronary Intervention in

89

Table 6: Life Time Cumulative Probability of Myocardial Infarction & PCI

Strategy Myocardial Infarction

(%) PCI

* (%)

Medical Rx 19.5 24.0

BMS 22.3 15.8

DES 22.9 12.3

* For BMS and DES strategies, this does not include the initial PCI.

Page 100: Economic Evaluation of Percutaneous Coronary Intervention in

90

Table 7: Base Case Results

Strategy Cost

Life-expectancy

(life-years)

Quality-Adjusted

Life-Expectancy

(QALY*)

ICER†

($/QALY)

Discounted

Medical Rx $22,952 10.86 10.10 -

BMS $25,081 10.99 10.26 $13,271

DES $25,536 10.94 10.20 Dominated

Non-discounted

Medical Rx $34,487 17.47 16.30 -

BMS $37,223 17.83 16.65 $7,365

DES $37,407 17.62 16.45 Dominated

* QALY is quality-adjusted life-years

† ICER is incremental cost effectiveness ratio

Page 101: Economic Evaluation of Percutaneous Coronary Intervention in

91

Table 8: Sub-group Analysis based on Risk of Restenosis

Strategy Cost QALY ICER

base-case

Medical Rx $22,953 10.10

BMS $25,081 10.26 $13,272

DES $25,536 10.20 (Dominated)

non-diabetic short lesion ( < 20 mm) and large artery (≥ 3mm)

Medical Rx $22,482 10.11

BMS $23,541 10.29 $5,814

DES $24,573 10.22 (Dominated)

non-diabetic long lesion (≥ 20 mm) and small artery (<3 mm)

Medical Rx $23,014 10.10

BMS $25,367 10.27 $14,414

DES $26,980 10.20 (Dominated)

diabetic short lesion ( < 20 mm) and large artery (≥ 3mm)

Medical Rx $25,266 10.07

BMS $28,029 10.21 $19,411

DES $28,363 10.18 (Dominated)

diabetic short lesion (< 20mm) and small artery (< 3mm)

Medical Rx $24,604 10.08

DES $26,827 10.19 $19,311

BMS $27,240 10.22 $14,345

diabetic long lesion (≥ 20 mm) and large artery (≥ 3mm)

Medical Rx $22,536 10.11

DES $23,566 10.23 $8,949

BMS $23,596 10.29 $463

diabetic long lesion (≥ 20 mm) and small artery (<3 mm)

Medical Rx $23,128 10.10

DES $25,135 10.21 $18,286

BMS $26,727 10.24 $61,985

Page 102: Economic Evaluation of Percutaneous Coronary Intervention in

92

Table 9: Scenario Analysis based on Initial Symptom Severity

Strategy Cost QALY ICER

CCS 0-1

Medical Rx $19,154 10.14

BMS $25,081 10.28 $43,739

DES $25,536 10.22 (Dominated)

CCS 2

Medical Rx $23,562 10.13

BMS $25,081 10.28 $10,584

DES $25,536 10.22 (Dominated)

CCS 3

Medical Rx $23,562 10.09

BMS $25,081 10.26 $8,914

DES $25,536 10.20 (Dominated)

Page 103: Economic Evaluation of Percutaneous Coronary Intervention in

93

Figure 1: Model Structure

Page 104: Economic Evaluation of Percutaneous Coronary Intervention in

94

Figure 2: Cost-Effectiveness Acceptability Curve

* refers to proportion of simulations in each strategy was the optimal decision, based on the corresponding willingness to pay

threshold.

Page 105: Economic Evaluation of Percutaneous Coronary Intervention in

95

5 Synthesis

In this thesis, we sought to address three major gaps in knowledge regarding the therapy of

patients with stable coronary artery disease. First, the degree of angina relief associated with

PCI as compared with medical therapy was estimated through a systematic review of the

literature and meta-analysis. Second, a mapping tool was created to transform SAQ quality of

life data from previous studies to utility weights for secondary use in economic evaluations.

Finally we created a comprehensive decision analytic model to evaluate the incremental cost-

effectiveness of PCI with either DES or BMS compared to aggressive medical therapy alone for

patients with stable coronary artery disease, calibrated to real world data, and incorporating all

sources of previous evidence.

5.1 Novel Findings

In chapter 2, we found that PCI when added to medical therapy was associated with an overall

improvement in angina relief compared with medical therapy alone for patients with stable

coronary artery disease. However, the incremental benefit of PCI on angina relief diminished

substantially over the trial periods included in our meta-analysis, with the benefit in angina relief

associated with PCI restricted predominantly to older trials. Contemporary studies did not show

any significant differences between patients treated with PCI and medical therapy. One key

reason might be improvement in the proportion of medically treated patients who became angina

free over time, from 39.9% in older trials to 56.7% in intermediate trials to 74.8% in

contemporary trials. The increasing proportion of angina free patients corresponded to improved

use of evidence-based medical therapy among trials. Indeed, we found that there was an inverse

relationship between utilization of evidence-based therapies and efficacy of PCI in our meta-

regression analysis. This chapter suggests that improvements in medical therapy, which in turn

have resulted in greater angina relief, may explain the attenuated incremental benefit of PCI on

in contemporary practice.

In chapter 3, we assessed several mapping algorithms to predict EQ-5D utility scores from

responses on the descriptive SAQ in a cohort with coronary artery disease. Although the best-

fitting model exhibited only modest predictive ability at the individual level, it was able to

Page 106: Economic Evaluation of Percutaneous Coronary Intervention in

96

accurately predict the mean EQ-5D utilities scores across a spectrum of symptom severity.

Using this tool, more up-to-date and symptom specific utility weights can be estimated for use in

economic evaluations.

Finally, in chapter 4, we present a well-validated state transition model based on Ontario

epidemiologic and cost data to compare medical therapy versus PCI for coronary artery disease.

We found that an initial strategy of PCI with a BMS stent was cost effective with an ICER of

$13,271 /QALY gained. This finding was consistent across the spectrum of clinically relevant

subgroups. DES was only cost-effective in patients with diabetes who had long lesions and

small arteries. Based on a probabilistic sensitivity analysis, we found substantial parameter

uncertainty, highlighting the need for further research to both understand the nature of this

uncertainty and further refine parameter estimates. Nonetheless, the results of our analysis,

based on the best current evidence, should aid both clinicians and policy makers in the initial

choice of therapy for patients with stable angina, who have undergone coronary angiography.

5.2 Implications for Clinical Practice

Due to a number of recent and highly controversial studies, there is significant debate as to the

appropriate initial strategy in patients with symptomatic stable angina and a documented

significant coronary artery lesion (8;9;141;158). This controversy has arisen because, in

comparison to medical therapy, PCI reduces angina but does not have a statistically significant

impact on mortality (13). This symptomatic improvement is limited by the phenomenon of

restenosis (3;13;145). Although DES are more efficacious than BMS in angina relief, there is

concern about long term safety and cost (158). Multiple previous economic evaluations have

concentrated on comparing BMS to DES stents, without incorporating a medical therapy arm

(54;62;71;72;159-162). In light of recent evidence that medical therapy may be equivalent to

PCI for many patient populations, this comparison is important.

Our comprehensive decision analytic model is a tool for evaluating these trade-offs and

potentially bringing clarity to this clinical controversy. This model found that PCI with a BMS

as an initial strategy is cost effective over medical therapy in most scenarios, except when the

risk of restenosis and recurrent symptoms is very high. This is in counter-distinction to current

practice. In Canada, DES are used in approximately 50% of PCI cases, while in the United

Page 107: Economic Evaluation of Percutaneous Coronary Intervention in

97

States DES use is 80-90%. Restricting an initial DES strategy to only patients at a high-risk for

restenosis will have substantial cost-savings.

Our target population was patients with documented angiographic coronary disease that was

deemed appropriate for PCI. This is similar to the patient population studied in clinical trials

such as the COURAGE trial. These conclusions must not be extended to all patients with

coronary artery disease, in particular patients who would not undergo coronary angiography, or

those with multi-vessel disease, in whom CABG is an alternative therapeutic option, and not one

that we considered in our analyses.

Moreover, our conclusions do not apply to patients prior to confirmation of hemodynamically

significant disease with angiography, nor did we study the relative merits of diagnostic

angiography versus non-invasive stress testing in making this diagnosis. Our results do suggest

that when discussing the need for a diagnostic angiogram with patients, the possibility of

subsequent PCI and the implications of stenting with a BMS or DES should also be covered.

Our study is not advocating same-sitting or ad hoc PCI, but rather highlights that a

comprehensive discussion regarding the consequences of a diagnostic angiogram, including

subsequent treatment options should be had.

We evaluated the cost-effectiveness of initial treatment strategy, and not specifically the

appropriateness. Appropriateness is an area of growing importance in cardiovascular disease,

with appropriateness criteria published for a number of diagnostic tools, including stress

echocardiography, cardiac computer tomography and magnetic resonance imaging (4;175-180).

An accepted operational definition of an appropriate procedure is one in which the expected

incremental benefit based on clinical judgment, exceeds the expected negative consequences by

a sufficiently wide margin such that the procedure is generally considered acceptable care

(4;175-180). The American College of Cardiology (ACC) and American Heart Association

(AHA) recently published a consensus document elaborating on an appropriateness rating scale

for revascularization, derived using a modified Delphi process (4). This scale can be used to rate

the appropriateness of a PCI or CABG based on patient and lesion characteristics, such as angina

severity, coronary anatomy, intensity of medical therapy and results of non-invasive imaging (4).

Page 108: Economic Evaluation of Percutaneous Coronary Intervention in

98

An implicit assumption of our analysis is that the target patient population studied was in fact

appropriate for revascularization. We assessed the cost-effectiveness of various initial treatment

strategies within this appropriate population. Importantly, we do not advocate the use of a cost-

effective, yet inappropriate intervention. For example, an asymptomatic patient on minimal

medications without a large area of ischemia on perfusion imaging would not be appropriate for

revascularization, based on the appropriateness criteria developed to date. Translation of our

results into clinical practice must incorporate appropriateness of the procedure, patient

preference and cost-effectiveness.

5.3 Foci for Future Research

5.3.1 Real World Outcomes

A key strength of our decision analytic model is that transition probabilities and costs were

derived from real world data. However, the real world data was restricted to the two PCI

strategies. Parameters for medical therapy strategy were obtained from randomized control trial

data, and as such are susceptible to external validity/generalizability issues. For example, the

high medical adherence rates used in the COURAGE trial, and used in our model may not reflect

real world clinical practice (20). This is a key limitation, given that the intensity of evidence-

based medication uptake has implication for the effectiveness of angina relief for PCI versus

medical therapy, as per our meta-regression analysis in Chapter 2.

Moreover, as we have discussed, one cannot extend our results to all patients with chronic stable

coronary disease, because to do so would require the assumption that the characteristics of

patients in the CCN PCI registry with stable angina are reflective of all patients with chronic

stable coronary disease, irrespective of treatment strategy. Given that the patients in the CCN

database by definition underwent coronary angiography, they are very likely to have greater

symptom severity, and possibly greater disease severity that the overall population of patients

with chronic coronary disease.

This highlights the need for better real world data on all patients with chronic stable coronary

artery disease, for example through optimizing the use of existing administrative data sources,

such as CIHI or OHIP. Algorithms for identifying medically treated or re-vascularized patients

Page 109: Economic Evaluation of Percutaneous Coronary Intervention in

99

with stable coronary disease should be developed and validated. Such algorithms exist for

accurately identifying hospitalized patients with MI or ACS; however, this represents only a

small fraction of the patients with coronary artery disease. For example, in 2005 16,640 patients

were admitted with an MI in Ontario, while approximately 300,000 patients were estimated to

have chronic stable angina, of whom only 5260 had a PCI (2). Accurate data on this much large

chronic angina population is lacking.

Once algorithms for identification of these patients are developed, they will have multiple

potential applications in health services research, including the longitudinal evaluation of real

world outcomes and costs as a function of medication use and revascularization modality. These

in turn can further refine the parameters in our decision model.

5.3.2 Utilities

We developed a mapping algorithm with the objective of making SAQ data from clinical trials

available for use in economic evaluations. In order to maximize the use of this tool, a systematic

appraisal of the literature on SAQ data for this patient population is necessary, such that utilities

can be calculated. This in turn will represent an off the shelf catalog of utilities for stable angina

patients with different functional class and will represent a valuable resource to further enrich

economic analyses.

These mapped mean EQ-5D values will represent a cross-sectional estimate of global quality of

life. An important area of future research is a greater appreciation of temporal trends in utility

weights for chronic coronary artery disease patients, as a function of changing angina severity

and functional class. Using a prospective cohort study design, with serial SAQ and EQ-5D

measurements in patients, we will be able to evaluate the relationship between changes in quality

of life as it relates to coronary disease (as measured with the SAQ) and global quality of life (as

measured in EQ-5D).

5.3.3 Mortality is the Key Parameter

Our a priori hypothesis was that the decision on initial treatment modality would be driven by

the magnitude to which subsequent revascularization procedures would be reduced by PCI.

Surprisingly, our model was critically sensitive to the relative mortality estimates for both

medical therapy and DES in comparison to BMS, and not sensitive to the frequency of

Page 110: Economic Evaluation of Percutaneous Coronary Intervention in

100

revascularization. This highlights a use of decision modeling to clarify the primary driver of a

clinical decision. Despite the large number of studies that have evaluated medical versus PCI in

stable angina, there remains a substantial uncertainty in the point estimates surrounding mortality

(13;155;163-165). Our work reinforces the need for further large studies in this area, such as the

pending ISCHEMIA (International Study of Comparative Health Effectiveness with Medical and

Invasive Approaches) trial. This National Institute of Health (NIH) funded international multi-

center randomized control trial is evaluating the comparative effectiveness of medical therapy

alone versus catheterization with revascularization if appropriate in stable patients with moderate

to severe ischemia.

Techniques such as expected value of perfect information and expected value of partial perfect

information can quantify the consequences of this decisional uncertainty (42). However, such

analyses are computational intensive as they involve a second outer loop of Monte-Carlo

simulations, in addition to the inner loop done for probabilistic sensitivity analyses (42). If 1000

inner and outer loops are planned, the 1,000,000 total iterations needed are likely to be

computational infeasible with large Markov models such as ours. The development of

methodologies to explore uncertainty in the context of large, complex models is important

methodological area for future study that follows from our work.

5.4 Conclusion

In this thesis, we evaluated multiple aspects surrounding the care of patients with stable coronary

artery disease. We have challenged conventional wisdom and provided new insight as to the

degree of angina relief afforded by PCI. We have explored methodological challenges in

mapping and provided a useful tool for deriving utility weights. Finally, through our

comprehensive decision analytic model, we have provided a flexible tool to model the natural

history of patients with stable coronary artery disease based on treatment modality. We believe

these are important contributions to this field of study and will be of substantial use in health

services research, and policy.

Page 111: Economic Evaluation of Percutaneous Coronary Intervention in

101

References

(1) Tu JV, Austin PC, Filate WA, Johansen HL, Brien SE, Pilote L, Alter DA. Outcomes

of acute myocardial infarction in Canada. Can J Cardiol 2003; 19:893-901.

(2) Wijeysundera HC, Machado M, Farahati F, Wang X, Witteman W, van d, V, Tu JV,

Lee DS, Goodman SG, Petrella R, O'Flaherty M, Krahn M, Capewell S. Association of

temporal trends in risk factors and treatment uptake with coronary heart disease

mortality, 1994-2005. JAMA 2010; 303:1841-7.

(3) Smith SC, Jr., Feldman TE, Hirshfeld JW, Jr., Jacobs AK, Kern MJ, King SB, III,

Morrison DA, O'neill WW, Schaff HV, Whitlow PL, Williams DO, Antman EM, Smith

SC, Jr., Adams CD, Anderson JL, Faxon DP, Fuster V, Halperin JL, Hiratzka LF, Hunt

SA, Jacobs AK, Nishimura R, Ornato JP, Page RL, Riegel B. ACC/AHA/SCAI 2005

guideline update for percutaneous coronary intervention: a report of the American

College of Cardiology/American Heart Association Task Force on Practice Guidelines

(ACC/AHA/SCAI Writing Committee to Update the 2001 Guidelines for Percutaneous

Coronary Intervention). J Am Coll Cardiol 2006; 47:e1-121.

(4) Patel MR, Dehmer GJ, Hirshfeld JW, Smith PK, Spertus JA.

ACCF/SCAI/STS/AATS/AHA/ASNC 2009 Appropriateness Criteria for Coronary

Revascularization. A Report of the American College of Cardiology Foundation

Appropriateness Criteria Task Force, Society for Cardiovascular Angiography and

Interventions, Society of Thoracic Surgeons, American Association for Thoracic

Surgery, American Heart Association, and the American Society of Nuclear

Cardiology. Circulation 2009.

(5) Cardiac Care Network (CCN) Annual Report 2009/2010. 2010.

(6) WRITING GROUP, Lloyd-Jones D, Adams RJ, Brown TM, Carnethon M, Dai S, De

Simone G, Ferguson TB, Ford E, Furie K, Gillespie C, Go A, Greenlund K, Haase N,

Hailpern S, Ho PM, Howard V, Kissela B, Kittner S, Lackland D, Lisabeth L, Marelli

Page 112: Economic Evaluation of Percutaneous Coronary Intervention in

102

A, McDermott MM, Meigs J, Mozaffarian D, Mussolino M, Nichol G, Roger VL,

Rosamond W, Sacco R, Sorlie P, Stafford R, Thom T, Wasserthiel-Smoller S, Wong

ND, Wylie-Rosett J, on behalf of the American Heart Association Statistics Committee

and Stroke Statistics Subcommittee. Heart Disease and Stroke Statistics--2010 Update:

A Report From the American Heart Association. Circulation 2010; 121:e46-215.

(7) Babapulle MN, Joseph L, Belisle P, Brophy JM, Eisenberg MJ. A hierarchical Bayesian

meta-analysis of randomised clinical trials of drug-eluting stents. Lancet 2004;

364:583-91.

(8) Spaulding C, Daemen J, Boersma E, Cutlip DE, Serruys PW. A pooled analysis of data

comparing sirolimus-eluting stents with bare-metal stents. N Engl J Med 2007;

356:989-97.

(9) Stone GW, Moses JW, Ellis SG, Schofer J, Dawkins KD, Morice MC, Colombo A,

Schampaert E, Grube E, Kirtane AJ, Cutlip DE, Fahy M, Pocock SJ, Mehran R, Leon

MB. Safety and efficacy of sirolimus- and paclitaxel-eluting coronary stents. N Engl J

Med 2007; 356:998-1008.

(10) Garg P, Cohen DJ, Gaziano T, Mauri L. Balancing the risks of restenosis and stent

thrombosis in bare-metal versus drug-eluting stents: results of a decision analytic

model. J Am Coll Cardiol 2008; 51:1844-53.

(11) Windecker S, Remondino A, Eberli FR, Juni P, Raber L, Wenaweser P, Togni M,

Billinger M, Tuller D, Seiler C, Roffi M, Corti R, Sutsch G, Maier W, Luscher T, Hess

OM, Egger M, Meier B. Sirolimus-eluting and paclitaxel-eluting stents for coronary

revascularization. N Engl J Med 2005; 353:653-62.

(12) Ko DT, Yun L, Wijeysundera HC, Jackevicius CA, Rao SV, Austin PC, Marquis JF, Tu

JV. Incidence, predictors, and prognostic implications of hospitalization for late

bleeding after percutaneous coronary intervention for patients older than 65 years. Circ

Cardiovasc Interv 2010; 3:140-7.

Page 113: Economic Evaluation of Percutaneous Coronary Intervention in

103

(13) Boden WE, O'Rourke RA, Teo KK, Hartigan PM, Maron DJ, Kostuk WJ, Knudtson M,

Dada M, Casperson P, Harris CL, Chaitman BR, Shaw L, Gosselin G, Nawaz S, Title

LM, Gau G, Blaustein AS, Booth DC, Bates ER, Spertus JA, Berman DS, Mancini GB,

Weintraub WS. Optimal medical therapy with or without PCI for stable coronary

disease. N Engl J Med 2007; 356:1503-16.

(14) Katritsis DG, Ioannidis JP. Percutaneous coronary intervention versus conservative

therapy in nonacute coronary artery disease: a meta-analysis. Circulation 2005;

111:2906-12.

(15) Schomig A, Mehilli J, de Waha A, Seyfarth M, Pache J, Kastrati A. A meta-analysis of

17 randomized trials of a percutaneous coronary intervention-based strategy in patients

with stable coronary artery disease. J Am Coll Cardiol 2008; 52:894-904.

(16) Henderson RA, Pocock SJ, Clayton TC, Knight R, Fox KA, Julian DG, Chamberlain

DA. Seven-year outcome in the RITA-2 trial: coronary angioplasty versus medical

therapy. J Am Coll Cardiol 2003; 42:1161-70.

(17) Bech GJ, De Bruyne B, Pijls NH, de Muinck ED, Hoorntje JC, Escaned J, Stella PR,

Boersma E, Bartunek J, Koolen JJ, Wijns W. Fractional flow reserve to determine the

appropriateness of angioplasty in moderate coronary stenosis: a randomized trial.

Circulation 2001; 103:2928-34.

(18) Hambrecht R, Walther C, Mobius-Winkler S, Gielen S, Linke A, Conradi K, Erbs S,

Kluge R, Kendziorra K, Sabri O, Sick P, Schuler G. Percutaneous coronary angioplasty

compared with exercise training in patients with stable coronary artery disease: a

randomized trial. Circulation 2004; 109:1371-8.

(19) Hochman JS, Lamas GA, Buller CE, Dzavik V, Reynolds HR, Abramsky SJ, Forman

S, Ruzyllo W, Maggioni AP, White H, Sadowski Z, Carvalho AC, Rankin JM, Renkin

JP, Steg PG, Mascette AM, Sopko G, Pfisterer ME, Leor J, Fridrich V, Mark DB,

Knatterud GL. Coronary intervention for persistent occlusion after myocardial

infarction. N Engl J Med 2006; 355:2395-407.

Page 114: Economic Evaluation of Percutaneous Coronary Intervention in

104

(20) Boden WE, O'Rourke RA, Teo KK, Hartigan PM, Maron DJ, Kostuk WJ, Knudtson M,

Dada M, Casperson P, Harris CL, Chaitman BR, Shaw L, Gosselin G, Nawaz S, Title

LM, Gau G, Blaustein AS, Booth DC, Bates ER, Spertus JA, Berman DS, Mancini GB,

Weintraub WS. Optimal medical therapy with or without PCI for stable coronary

disease. N Engl J Med 2007; 356:1503-16.

(21) Coronary angioplasty versus medical therapy for angina: the second Randomised

Intervention Treatment of Angina (RITA-2) trial. RITA-2 trial participants. Lancet

1997; 350:461-8.

(22) Weintraub WS, Spertus JA, Kolm P, Maron DJ, Zhang Z, Jurkovitz C, Zhang W,

Hartigan PM, Lewis C, Veledar E, Bowen J, Dunbar SB, Deaton C, Kaufman S,

O'Rourke RA, Goeree R, Barnett PG, Teo KK, Boden WE, Mancini GB. Effect of PCI

on quality of life in patients with stable coronary disease. N Engl J Med 2008; 359:677-

87.

(23) Egger M, Smith GD, Altman DG. Systematic Reviews in Health Care; Meta-analysis in

context. 1-1-2001. London, BMJ Publishing Group.

(24) Wijeysundera HC, Ko DT. Does Percutaneous Coronary Intervention Reduce Mortality

in Patients With Stable Chronic Angina: Are We Talking About Apples and Oranges?

Circ Cardiovasc Qual Outcomes 2009; 2:123-6.

(25) Anderson JL, Adams CD, Antman EM, Bridges CR, Califf RM, Casey DE, Jr., Chavey

WE, Fesmire FM, Hochman JS, Levin TN, Lincoff AM, Peterson ED, Theroux P,

Wenger NK, Wright RS, Smith SC, Jr., Jacobs AK, Adams CD, Anderson JL, Antman

EM, Halperin JL, Hunt SA, Krumholz HM, Kushner FG, Lytle BW, Nishimura R,

Ornato JP, Page RL, Riegel B. ACC/AHA 2007 guidelines for the management of

patients with unstable angina/non-ST-Elevation myocardial infarction: a report of the

American College of Cardiology/American Heart Association Task Force on Practice

Guidelines (Writing Committee to Revise the 2002 Guidelines for the Management of

Patients With Unstable Angina/Non-ST-Elevation Myocardial Infarction) developed in

collaboration with the American College of Emergency Physicians, the Society for

Page 115: Economic Evaluation of Percutaneous Coronary Intervention in

105

Cardiovascular Angiography and Interventions, and the Society of Thoracic Surgeons

endorsed by the American Association of Cardiovascular and Pulmonary Rehabilitation

and the Society for Academic Emergency Medicine. J Am Coll Cardiol 2007; 50:e1-

e157.

(26) Keeley EC, Boura JA, Grines CL. Primary angioplasty versus intravenous thrombolytic

therapy for acute myocardial infarction: a quantitative review of 23 randomised trials.

Lancet 2003; 361:13-20.

(27) Wijeysundera HC, You JJ, Nallamothu BK, Krumholz HM, Cantor WJ, Ko DT. An

early invasive strategy versus ischemia-guided management after fibrinolytic therapy

for ST-segment elevation myocardial infarction: a meta-analysis of contemporary

randomized controlled trials. Am Heart J 2008; 156:564-72, 572.

(28) Huedo-Medina TB, Sanchez-Meca J, Marin-Martinez F, Botella J. Assessing

heterogeneity in meta-analysis: Q statistic or I2 index? Psychol Methods 2006; 11:193-

206.

(29) Fryback DG, Dasbach EJ, Klein R, Klein BE, Dorn N, Peterson K, Martin PA. The

Beaver Dam Health Outcomes Study: initial catalog of health-state quality factors. Med

Decis Making 1993; 13:89-102.

(30) Fryback DG, Lawrence WF, Martin PA, Klein R, Klein BE. Predicting Quality of

Well-being scores from the SF-36: results from the Beaver Dam Health Outcomes

Study. Med Decis Making 1997; 17:1-9.

(31) Llewellyn-Thomas HA. Health state descriptions. Purposes, issues, a proposal. Med

Care 1996; 34:DS109-DS118.

(32) Mortimer D, Segal L. Comparing the incomparable? A systematic review of competing

techniques for converting descriptive measures of health status into QALY-weights.

Med Decis Making 2008; 28:66-89.

Page 116: Economic Evaluation of Percutaneous Coronary Intervention in

106

(33) Brooks R. EuroQol: the current state of play. Health Policy 1996; 37:53-72.

(34) Dolan P, Roberts J. Modelling valuations for Eq-5d health states: an alternative model

using differences in valuations. Med Care 2002; 40:442-6.

(35) Lenert L, Kaplan RM. Validity and interpretation of preference-based measures of

health-related quality of life. Med Care 2000; 38:II138-II150.

(36) Coons SJ, Rao S, Keininger DL, Hays RD. A comparative review of generic quality-of-

life instruments. Pharmacoeconomics 2000; 17:13-35.

(37) Garratt AM, Hutchinson A, Russell I. The UK version of the Seattle Angina

Questionnaire (SAQ-UK): reliability, validity and responsiveness. J Clin Epidemiol

2001; 54:907-15.

(38) Spertus JA, Winder JA, Dewhurst TA, Deyo RA, Prodzinski J, McDonell M, Fihn SD.

Development and evaluation of the Seattle Angina Questionnaire: a new functional

status measure for coronary artery disease. J Am Coll Cardiol 1995; 25:333-41.

(39) Longworth L, Buxton MJ, Sculpher M, Smith DH. Estimating utility data from clinical

indicators for patients with stable angina. Eur J Health Econ 2005; 6:347-53.

(40) National Institute for Clinical Excellence (2008). Guide to the methods of technology

appraisals. Available at http://www.nice.org.uk; accessed November 2009. 2009.

(41) Hunink M, Glasziou P, Siegel J, Weeks J, Pliskin J, Elstein A, Weinstein M. Decision

Making in health and medicine. 1-1-2001. Cambridge, Cambridge University Press.

(42) Briggs A, Claxton K, Sculpher M. Decision Modelling for Health Economic

Evaluation. 1-1-2006. Oxford, Oxford University Press.

(43) Drummond MF, Sculpher MJ, Torrance GW, O'Brien BJ, Stoddart GL. Methods for the

Economic Evaluation of Health Care Programmes. 1-1-1987. Oxford, Oxford

University Press.

Page 117: Economic Evaluation of Percutaneous Coronary Intervention in

107

(44) Detsky AS, Naglie G, Krahn MD, Redelmeier DA, Naimark D. Primer on medical

decision analysis: Part 2--Building a tree. Med Decis Making 1997; 17:126-35.

(45) Detsky AS, Naglie G, Krahn MD, Naimark D, Redelmeier DA. Primer on medical

decision analysis: Part 1--Getting started. Med Decis Making 1997; 17:123-5.

(46) Krahn MD, Naglie G, Naimark D, Redelmeier DA, Detsky AS. Primer on medical

decision analysis: Part 4--Analyzing the model and interpreting the results. Med Decis

Making 1997; 17:142-51.

(47) Naimark D, Krahn MD, Naglie G, Redelmeier DA, Detsky AS. Primer on medical

decision analysis: Part 5--Working with Markov processes. Med Decis Making 1997;

17:152-9.

(48) Claude J, Schindler C, Kuster GM, Schwenkglenks M, Szucs T, Buser P, Osswald S,

Kaiser C, Gradel C, Estlinbaum W, Rickenbacher P, Pfisterer M. Cost-effectiveness of

invasive versus medical management of elderly patients with chronic symptomatic

coronary artery disease. Findings of the randomized trial of invasive versus medical

therapy in elderly patients with chronic angina (TIME). Eur Heart J 2004; 25:2195-203.

(49) Griffin SC, Barber JA, Manca A, Sculpher MJ, Thompson SG, Buxton MJ,

Hemingway H. Cost effectiveness of clinically appropriate decisions on alternative

treatments for angina pectoris: prospective observational study. BMJ 2007; 334:624.

(50) Hlatky MA, Boothroyd DB, Melsop KA, Kennedy L, Rihal C, Rogers WJ,

Venkitachalam L, Brooks MM. Economic outcomes of treatment strategies for type 2

diabetes mellitus and coronary artery disease in the Bypass Angioplasty

Revascularization Investigation 2 Diabetes trial. Circulation 2009; 120:2550-8.

(51) Weintraub WS, Boden WE, Zhang Z, Kolm P, Zhang Z, Spertus JA, Hartigan P,

Veledar E, Jurkovitz C, Bowen J, Maron DJ, O'Rourke R, Dada M, Teo KK, Goeree R,

Barnett PG. Cost-effectiveness of percutaneous coronary intervention in optimally

treated stable coronary patients. Circ Cardiovasc Qual Outcomes 2008; 1:12-20.

Page 118: Economic Evaluation of Percutaneous Coronary Intervention in

108

(52) Byrne RA, Sarafoff N, Kastrati A, Schomig A. Drug-eluting stents in percutaneous

coronary intervention: a benefit-risk assessment. Drug Saf 2009; 32:749-70.

(53) Cleland JG, Walker A. Is medical treatment for angina the most cost-effective option?

Eur Heart J 1997; 18 Suppl B:B35-B42.

(54) Eisenberg MJ. Drug-eluting stents: the price is not right. Circulation 2006; 114:1745-

54.

(55) Groeneveld PW, Suh JJ, Matta MA. The costs and quality-of-life outcomes of drug-

eluting coronary stents: a systematic review. J Interv Cardiol 2007; 20:1-9.

(56) Hill R, Bagust A, Bakhai A, Dickson R, Dundar Y, Haycox A, Mujica MR, Reaney A,

Roberts D, Williamson P, Walley T. Coronary artery stents: a rapid systematic review

and economic evaluation. Health Technol Assess 2004; 8:iii-242.

(57) Kirtane AJ, Cohen DJ. When is better not good enough?: Insights from the COURAGE

economic study. Circ Cardiovasc Qual Outcomes 2008; 1:4-6.

(58) Kong DF, Eisenstein EL. Decision models for assessing the cost effectiveness of drug-

eluting stents. Expert Opin Pharmacother 2005; 6:965-74.

(59) Ligthart S, Vlemmix F, Dendukuri N, Brophy JM. The cost-effectiveness of drug-

eluting stents: a systematic review. CMAJ 2007; 176:199-205.

(60) Lord SJ, Howard K, Allen F, Marinovich L, Burgess DC, King R, Atherton JJ. A

systematic review and economic analysis of drug-eluting coronary stents available in

Australia. Med J Aust 2005; 183:464-71.

(61) O'Rourke RA. Cost-effective management of chronic stable angina. Clin Cardiol 1996;

19:497-501.

Page 119: Economic Evaluation of Percutaneous Coronary Intervention in

109

(62) Ryan J, Cohen DJ. Are drug-eluting stents cost-effective? It depends on whom you ask.

Circulation 2006; 114:1736-43.

(63) Toutouzas K, Synetos A, Karanasos A, Drakopoulou M, Tsiamis E, Lerakis S,

Stefanadis C. Percutaneous coronary intervention in chronic stable angina. Am J Med

Sci 2010; 339:568-72.

(64) Sculpher MJ, Petticrew M, Kelland JL, Elliott RA, Holdright DR, Buxton MJ.

Resource allocation for chronic stable angina: a systematic review of effectiveness,

costs and cost-effectiveness of alternative interventions. Health Technol Assess 1998;

2:i-176.

(65) Leon MB, Kandzari DE, Eisenstein EL, Anstrom KJ, Mauri L, Cutlip DE, Nikolsky E,

O'Shaughnessy C, Overlie PA, Kirtane AJ, McLaurin BT, Solomon SL, Douglas JS, Jr.,

Popma JJ. Late safety, efficacy, and cost-effectiveness of a zotarolimus-eluting stent

compared with a paclitaxel-eluting stent in patients with de novo coronary lesions: 2-

year follow-up from the ENDEAVOR IV trial (Randomized, Controlled Trial of the

Medtronic Endeavor Drug [ABT-578] Eluting Coronary Stent System Versus the Taxus

Paclitaxel-Eluting Coronary Stent System in De Novo Native Coronary Artery

Lesions). JACC Cardiovasc Interv 2009; 2:1208-18.

(66) Eisenstein EL, Leon MB, Kandzari DE, Mauri L, Edwards R, Kong DF, Cowper PA,

Anstrom KJ. Long-term clinical and economic analysis of the Endeavor zotarolimus-

eluting stent versus the cypher sirolimus-eluting stent: 3-year results from the

ENDEAVOR III trial (Randomized Controlled Trial of the Medtronic Endeavor Drug

[ABT-578] Eluting Coronary Stent System Versus the Cypher Sirolimus-Eluting

Coronary Stent System in De Novo Native Coronary Artery Lesions). JACC

Cardiovasc Interv 2009; 2:1199-207.

(67) Bagust A, Grayson AD, Palmer ND, Perry RA, Walley T. Cost effectiveness of drug

eluting coronary artery stenting in a UK setting: cost-utility study. Heart 2006; 92:68-

74.

Page 120: Economic Evaluation of Percutaneous Coronary Intervention in

110

(68) Bakhai A, Stone GW, Mahoney E, Lavelle TA, Shi C, Berezin RH, Lahue BJ, Clark

MA, Lacey MJ, Russell ME, Ellis SG, Hermiller JB, Cox DA, Cohen DJ. Cost

effectiveness of paclitaxel-eluting stents for patients undergoing percutaneous coronary

revascularization: results from the TAXUS-IV Trial. J Am Coll Cardiol 2006; 48:253-

61.

(69) Brunner-La Rocca HP, Kaiser C, Bernheim A, Zellweger MJ, Jeger R, Buser PT,

Osswald S, Pfisterer M. Cost-effectiveness of drug-eluting stents in patients at high or

low risk of major cardiac events in the Basel Stent KostenEffektivitats Trial

(BASKET): an 18-month analysis. Lancet 2007; 370:1552-9.

(70) Cohen DJ, Bakhai A, Shi C, Githiora L, Lavelle T, Berezin RH, Leon MB, Moses JW,

Carrozza JP, Jr., Zidar JP, Kuntz RE. Cost-effectiveness of sirolimus-eluting stents for

treatment of complex coronary stenoses: results from the Sirolimus-Eluting Balloon

Expandable Stent in the Treatment of Patients With De Novo Native Coronary Artery

Lesions (SIRIUS) trial. Circulation 2004; 110:508-14.

(71) Rinfret S, Cohen DJ, Tahami Monfared AA, Lelorier J, Mireault J, Schampaert E. Cost

effectiveness of the sirolimus-eluting stent in high-risk patients in Canada: an analysis

from the C-SIRIUS trial. Am J Cardiovasc Drugs 2006; 6:159-68.

(72) Shrive FM, Manns BJ, Galbraith PD, Knudtson ML, Ghali WA. Economic evaluation

of sirolimus-eluting stents. CMAJ 2005; 172:345-51.

(73) Goeree R, Bowen JM, Blackhouse G, Lazzam C, Cohen E, Chiu M, Hopkins R, Tarride

JE, Tu JV. Economic evaluation of drug-eluting stents compared to bare metal stents

using a large prospective study in Ontario. Int J Technol Assess Health Care 2009;

25:196-207.

(74) Ikeda S, Kobayashi M. Economic evaluation of drug-eluting stents in Japan. Keio J

Med 2006; 55:15-22.

Page 121: Economic Evaluation of Percutaneous Coronary Intervention in

111

(75) Polanczyk CA, Wainstein MV, Ribeiro JP. Cost-effectiveness of sirolimus-eluting

stents in percutaneous coronary interventions in Brazil. Arq Bras Cardiol 2007; 88:464-

74.

(76) Mahieu J, De RA, De GD, Vrints C, Bosmans J. Economic analysis of the use of drug-

eluting stents from the perspective of Belgian health care. Acta Cardiol 2007; 62:355-

65.

(77) Tamburino C, Barbagallo R, Capodanno D, di MS, Colombo GL, Recchia M,

Ciriminna S. Cost-effectiveness of the real-world use of drug-eluting stents at 9-month

follow-up: results from the Sicilian DES Registry. J Cardiovasc Med (Hagerstown )

2009; 10:322-9.

(78) Ekman M, Sjogren I, James S. Cost-effectiveness of the Taxus paclitaxel-eluting stent

in the Swedish healthcare system. Scand Cardiovasc J 2006; 40:17-24.

(79) Varani E, Guastaroba P, Di Tanna GL, Saia F, Balducelli M, Campo G, Vignali L,

Rossi R, Manari A, Piovaccari G, De PR, Marzocchi A. Long-term clinical outcomes

and cost-effectiveness analysis in multivessel percutaneous coronary interventions:

comparison of drug-eluting stents, bare-metal stents and a mixed approach in patients at

high and low risk of repeat revascularisation. EuroIntervention 2010; 5:953-61.

(80) Remak E, Manson S, Hutton J, Brasseur P, Olivier E, Gershlick A. Cost-effectiveness

of the Endeavor stent in de novo native coronary artery lesions updated with

contemporary data. EuroIntervention 2010; 5:826-32.

(81) Eisenstein EL, Wijns W, Fajadet J, Mauri L, Edwards R, Cowper PA, Kong DF,

Anstrom KJ. Long-term clinical and economic analysis of the Endeavor drug-eluting

stent versus the Driver bare-metal stent: 4-year results from the ENDEAVOR II trial

(Randomized Controlled Trial to Evaluate the Safety and Efficacy of the Medtronic

AVE ABT-578 Eluting Driver Coronary Stent in De Novo Native Coronary Artery

Lesions). JACC Cardiovasc Interv 2009; 2:1178-87.

Page 122: Economic Evaluation of Percutaneous Coronary Intervention in

112

(82) Wong JB, Sonnenberg FA, Salem DN, Pauker SG. Myocardial revascularization for

chronic stable angina. Analysis of the role of percutaneous transluminal coronary

angioplasty based on data available in 1989. Ann Intern Med 1990; 113:852-71.

(83) Sculpher M, Smith D, Clayton T, Henderson R, Buxton M, Pocock S, Chamberlain D.

Coronary angioplasty versus medical therapy for angina. Health service costs based on

the second Randomized Intervention Treatment of Angina (RITA-2) trial. Eur Heart J

2002; 23:1291-300.

(84) Favarato D, Hueb W, Gersh BJ, Soares PR, Cesar LA, da Luz PL, Oliveira SA,

Ramires JA. Relative cost comparison of treatments for coronary artery disease: the

First Year Follow-Up of MASS II Study. Circulation 2003; 108 Suppl 1:II21-II23.

(85) Weintraub WS, Barnett P, Chen S, Hartigan P, Casperson P, O'Rourke R, Boden WE,

Lewis C, Veledar E, Becker E, Culler S, Kolm P, Mahoney EM, Dunbar SB, Deaton C,

O'Brien B, Goeree R, Blackhouse G, Nease R, Spertus J, Kaufman S, Teo K.

Economics methods in the Clinical Outcomes Utilizing percutaneous coronary

Revascularization and Aggressive Guideline-driven drug Evaluation (COURAGE) trial.

Am Heart J 2006; 151:1180-5.

(86) Mittmann N, Trakas K, Risebrough N, Liu BA. Utility scores for chronic conditions in

a community-dwelling population. Pharmacoeconomics 1999; 15:369-76.

(87) Nease RF, Jr., Kneeland T, O'Connor GT, Sumner W, Lumpkins C, Shaw L, Pryor D,

Sox HC. Variation in patient utilities for outcomes of the management of chronic stable

angina. Implications for clinical practice guidelines. Ischemic Heart Disease Patient

Outcomes Research Team. JAMA 1995; 273:1185-90.

(88) Rosamond W, Flegal K, Furie K, Go A, Greenlund K, Haase N, Hailpern SM, Ho M,

Howard V, Kissela B, Kittner S, Lloyd-Jones D, McDermott M, Meigs J, Moy C,

Nichol G, O'Donnell C, Roger V, Sorlie P, Steinberger J, Thom T, Wilson M, Hong Y.

Heart disease and stroke statistics--2008 update: a report from the American Heart

Page 123: Economic Evaluation of Percutaneous Coronary Intervention in

113

Association Statistics Committee and Stroke Statistics Subcommittee. Circulation 2008;

117:e25-146.

(89) Stroup DF, Berlin JA, Morton SC, Olkin I, Williamson GD, Rennie D, Moher D,

Becker BJ, Sipe TA, Thacker SB. Meta-analysis of observational studies in

epidemiology: a proposal for reporting. Meta-analysis Of Observational Studies in

Epidemiology (MOOSE) group. JAMA 2000; 283:2008-12.

(90) Erne P, Schoenenberger AW, Burckhardt D, Zuber M, Kiowski W, Buser PT, Dubach

P, Resink TJ, Pfisterer M. Effects of percutaneous coronary interventions in silent

ischemia after myocardial infarction: the SWISSI II randomized controlled trial. JAMA

2007; 297:1985-91.

(91) Dakik HA, Kleiman NS, Farmer JA, He ZX, Wendt JA, Pratt CM, Verani MS,

Mahmarian JJ. Intensive medical therapy versus coronary angioplasty for suppression

of myocardial ischemia in survivors of acute myocardial infarction: a prospective,

randomized pilot study. Circulation 1998; 98:2017-23.

(92) Ellis SG, Mooney MR, George BS, da Silva EE, Talley JD, Flanagan WH, Topol EJ.

Randomized trial of late elective angioplasty versus conservative management for

patients with residual stenoses after thrombolytic treatment of myocardial infarction.

Treatment of Post-Thrombolytic Stenoses (TOPS) Study Group. Circulation 1992;

86:1400-6.

(93) Zeymer U, Uebis R, Vogt A, Glunz HG, Vohringer HF, Harmjanz D, Neuhaus KL.

Randomized comparison of percutaneous transluminal coronary angioplasty and

medical therapy in stable survivors of acute myocardial infarction with single vessel

disease: a study of the Arbeitsgemeinschaft Leitende Kardiologische Krankenhausarzte.

Circulation 2003; 108:1324-8.

(94) Steg PG, Thuaire C, Himbert D, Carrie D, Champagne S, Coisne D, Khalife K, Cazaux

P, Logeart D, Slama M, Spaulding C, Cohen A, Tirouvanziam A, Montely JM,

Rodriguez RM, Garbarz E, Wijns W, Durand-Zaleski I, Porcher R, Brucker L, Chevret

Page 124: Economic Evaluation of Percutaneous Coronary Intervention in

114

S, Chastang C. DECOPI (DEsobstruction COronaire en Post-Infarctus): a randomized

multi-centre trial of occluded artery angioplasty after acute myocardial infarction. Eur

Heart J 2004; 25:2187-94.

(95) Jeremias A, Kaul S, Rosengart TK, Gruberg L, Brown DL. The impact of

revascularization on mortality in patients with nonacute coronary artery disease. Am J

Med 2009; 122:152-61.

(96) Jadad AR, Moore RA, Carroll D, Jenkinson C, Reynolds DJ, Gavaghan DJ, McQuay

HJ. Assessing the quality of reports of randomized clinical trials: is blinding necessary?

Control Clin Trials 1996; 17:1-12.

(97) Hueb WA, Bellotti G, de Oliveira SA, Arie S, de Albuquerque CP, Jatene AD, Pileggi

F. The Medicine, Angioplasty or Surgery Study (MASS): a prospective, randomized

trial of medical therapy, balloon angioplasty or bypass surgery for single proximal left

anterior descending artery stenoses. J Am Coll Cardiol 1995; 26:1600-5.

(98) Hueb WA, Soares PR, Almeida DO, Arie S, Cardoso RH, Wajsbrot DB, Cesar LA,

Jatene AD, Ramires JA. Five-year follow-op of the medicine, angioplasty, or surgery

study (MASS): A prospective, randomized trial of medical therapy, balloon

angioplasty, or bypass surgery for single proximal left anterior descending coronary

artery stenosis. Circulation 1999; 100:II107-II113.

(99) Huedo-Medina TB, Sanchez-Meca J, Marin-Martinez F, Botella J. Assessing

heterogeneity in meta-analysis: Q statistic or I2 index? Psychol Methods 2006; 11:193-

206.

(100) Berkey CS, Hoaglin DC, Mosteller F, Colditz GA. A random-effects regression model

for meta-analysis. Stat Med 1995; 14:395-411.

(101) Thompson SG, Sharp SJ. Explaining heterogeneity in meta-analysis: a comparison of

methods. Stat Med 1999; 18:2693-708.

Page 125: Economic Evaluation of Percutaneous Coronary Intervention in

115

(102) Thompson SG, Higgins JP. How should meta-regression analyses be undertaken and

interpreted? Stat Med 2002; 21:1559-73.

(103) Morton SC AJSMSPG. Meta-regression Approaches: What, Why,When, and How?

AHRQ Publication No 04-0033 2004.

(104) Trial of invasive versus medical therapy in elderly patients with chronic symptomatic

coronary-artery disease (TIME): a randomised trial. Lancet 2001; 358:951-7.

(105) Folland ED, Hartigan PM, Parisi AF. Percutaneous transluminal coronary angioplasty

versus medical therapy for stable angina pectoris: outcomes for patients with double-

vessel versus single-vessel coronary artery disease in a Veterans Affairs Cooperative

randomized trial. Veterans Affairs ACME InvestigatorS. J Am Coll Cardiol 1997;

29:1505-11.

(106) Hartigan PM, Giacomini JC, Folland ED, Parisi AF. Two- to three-year follow-up of

patients with single-vessel coronary artery disease randomized to PTCA or medical

therapy (results of a VA cooperative study). Veterans Affairs Cooperative Studies

Program ACME Investigators. Angioplasty Compared to Medicine. Am J Cardiol

1998; 82:1445-50.

(107) Hueb W, Soares PR, Gersh BJ, Cesar LA, Luz PL, Puig LB, Martinez EM, Oliveira

SA, Ramires JA. The medicine, angioplasty, or surgery study (MASS-II): a

randomized, controlled clinical trial of three therapeutic strategies for multivessel

coronary artery disease: one-year results. J Am Coll Cardiol 2004; 43:1743-51.

(108) Hueb W, Lopes NH, Gersh BJ, Soares P, Machado LA, Jatene FB, Oliveira SA,

Ramires JA. Five-year follow-up of the Medicine, Angioplasty, or Surgery Study

(MASS II): a randomized controlled clinical trial of 3 therapeutic strategies for

multivessel coronary artery disease. Circulation 2007; 115:1082-9.

(109) Pitt B, Waters D, Brown WV, van Boven AJ, Schwartz L, Title LM, Eisenberg D,

Shurzinske L, McCormick LS. Aggressive lipid-lowering therapy compared with

Page 126: Economic Evaluation of Percutaneous Coronary Intervention in

116

angioplasty in stable coronary artery disease. Atorvastatin versus Revascularization

Treatment Investigators. N Engl J Med 1999; 341:70-6.

(110) Madsen JK, Nielsen TT, Grande P, Eriksen UH, Saunamaki K, Thayssen P, Kassis E,

Rasmussen K, Haunso S, Haghfelt T, Fritz-Hansen P, Hjelms E, Paulsen PK, Alstrup P,

Arendrup H, Niebuhr-Jorgensen U, Andersen LI. Revascularization compared to

medical treatment in patients with silent vs. symptomatic residual ischemia after

thrombolyzed myocardial infarction--the DANAMI study. Cardiology 2007; 108:243-

51.

(111) Mahmarian JJ, Dakik HA, Filipchuk NG, Shaw LJ, Iskander SS, Ruddy TD, Keng F,

Henzlova MJ, Allam A, Moye LA, Pratt CM. An initial strategy of intensive medical

therapy is comparable to that of coronary revascularization for suppression of

scintigraphic ischemia in high-risk but stable survivors of acute myocardial infarction. J

Am Coll Cardiol 2006; 48:2458-67.

(112) Davies RF, Goldberg AD, Forman S, Pepine CJ, Knatterud GL, Geller N, Sopko G,

Pratt C, Deanfield J, Conti CR. Asymptomatic Cardiac Ischemia Pilot (ACIP) study

two-year follow-up: outcomes of patients randomized to initial strategies of medical

therapy versus revascularization. Circulation 1997; 95:2037-43.

(113) Dzavik V, Beanlands DS, Davies RF, Leddy D, Marquis JF, Teo KK, Ruddy TD,

Burton JR, Humen DP. Effects of late percutaneous transluminal coronary angioplasty

of an occluded infarct-related coronary artery on left ventricular function in patients

with a recent (< 6 weeks) Q-wave acute myocardial infarction (Total Occlusion Post-

Myocardial Infarction Intervention Study [TOMIIS]--a pilot study). Am J Cardiol 1994;

73:856-61.

(114) Horie H, Takahashi M, Minai K, Izumi M, Takaoka A, Nozawa M, Yokohama H,

Fujita T, Sakamoto T, Kito O, Okamura H, Kinoshita M. Long-term beneficial effect of

late reperfusion for acute anterior myocardial infarction with percutaneous transluminal

coronary angioplasty. Circulation 1998; 98:2377-82.

Page 127: Economic Evaluation of Percutaneous Coronary Intervention in

117

(115) Yousef ZR, Redwood SR, Bucknall CA, Sulke AN, Marber MS. Late intervention after

anterior myocardial infarction: effects on left ventricular size, function, quality of life,

and exercise tolerance: results of the Open Artery Trial (TOAT Study). J Am Coll

Cardiol 2002; 40:869-76.

(116) Pijls NH, van Schaardenburgh P, Manoharan G, Boersma E, Bech JW, van't Veer M,

Bar F, Hoorntje J, Koolen J, Wijns W, De Bruyne B. Percutaneous coronary

intervention of functionally nonsignificant stenosis: 5-year follow-up of the DEFER

Study. J Am Coll Cardiol 2007; 49:2105-11.

(117) A Randomized Trial of Therapies for Type 2 Diabetes and Coronary Artery Disease. N

Engl J Med 2009.

(118) King SB, III, Smith SC, Jr., Hirshfeld JW, Jr., Jacobs AK, Morrison DA, Williams DO,

Feldman TE, Kern MJ, O'neill WW, Schaff HV, Whitlow PL, Adams CD, Anderson

JL, Buller CE, Creager MA, Ettinger SM, Halperin JL, Hunt SA, Krumholz HM,

Kushner FG, Lytle BW, Nishimura R, Page RL, Riegel B, Tarkington LG, Yancy CW.

2007 focused update of the ACC/AHA/SCAI 2005 guideline update for percutaneous

coronary intervention: a report of the American College of Cardiology/American Heart

Association Task Force on Practice guidelines. J Am Coll Cardiol 2008; 51:172-209.

(119) Smith SC, Jr., Feldman TE, Hirshfeld JW, Jr., Jacobs AK, Kern MJ, King SB, III,

Morrison DA, O'Neill WW, Schaff HV, Whitlow PL, Williams DO, Antman EM,

Smith SC, Jr., Adams CD, Anderson JL, Faxon DP, Fuster V, Halperin JL, Hiratzka

LF, Hunt SA, Jacobs AK, Nishimura R, Ornato JP, Page RL, Riegel B.

ACC/AHA/SCAI 2005 guideline update for percutaneous coronary intervention: a

report of the American College of Cardiology/American Heart Association Task Force

on Practice Guidelines (ACC/AHA/SCAI Writing Committee to Update the 2001

Guidelines for Percutaneous Coronary Intervention). J Am Coll Cardiol 2006; 47:e1-

121.

Page 128: Economic Evaluation of Percutaneous Coronary Intervention in

118

(120) Patel MR, Dehmer GJ, Hirshfeld JW, Smith PK, Spertus JA.

ACCF/SCAI/STS/AATS/AHA/ASNC 2009 Appropriateness Criteria for Coronary

Revascularization: A Report of the American College of Cardiology Foundation

Appropriateness Criteria Task Force, Society for Cardiovascular Angiography and

Interventions, Society of Thoracic Surgeons, American Association for Thoracic

Surgery, American Heart Association, and the American Society of Nuclear

Cardiology: Endorsed by the American Society of Echocardiography, the Heart Failure

Society of America, and the Society of Cardiovascular Computed Tomography.

Circulation 2009; 119:1330-52.

(121) Guidelines for the economic evaluation of health technologies: Canada [3rd Edition].

Ottawa: Canadian Agency for Drugs and Technologies in Health; 2006. 2009.

(122) Spertus J, Decker C, Woodman C, House J, Jones P, O'Keefe J, Borkon AM. Effect of

difficulty affording health care on health status after coronary revascularization.

Circulation 2005; 111:2572-8.

(123) Spertus JA, Winder JA, Dewhurst TA, Deyo RA, Fihn SD. Monitoring the quality of

life in patients with coronary artery disease. Am J Cardiol 1994; 74:1240-4.

(124) Spertus JA, Bliven BD, Farner M, Gillen A, Hewitt T, Jones P, McCallister BD.

Integrating baseline health status data collection into the process of care. Jt Comm J

Qual Improv 2001; 27:369-80.

(125) Spertus JA, Jones P, McDonell M, Fan V, Fihn SD. Health status predicts long-term

outcome in outpatients with coronary disease. Circulation 2002; 106:43-9.

(126) Spertus JA, Salisbury AC, Jones PG, Conaway DG, Thompson RC. Predictors of

quality-of-life benefit after percutaneous coronary intervention. Circulation 2004;

110:3789-94.

(127) Spertus JA, Nerella R, Kettlekamp R, House J, Marso S, Borkon AM, Rumsfeld JS.

Risk of restenosis and health status outcomes for patients undergoing percutaneous

Page 129: Economic Evaluation of Percutaneous Coronary Intervention in

119

coronary intervention versus coronary artery bypass graft surgery. Circulation 2005;

111:768-73.

(128) Graham MM, Norris CM, Galbraith PD, Knudtson ML, Ghali WA. Quality of life after

coronary revascularization in the elderly. Eur Heart J 2006; 27:1690-8.

(129) Norris CM, Saunders LD, Ghali WA, Brant R, Galbraith PD, Graham M, Faris P,

Dzavik V, Knudtson ML. Health-related quality of life outcomes of patients with

coronary artery disease treated with cardiac surgery, percutaneous coronary

intervention or medical management. Can J Cardiol 2004; 20:1259-66.

(130) Briggs A, Clark T, Wolstenholme J, Clarke P. Missing... presumed at random: cost-

analysis of incomplete data. Health Econ 2003; 12:377-92.

(131) Austin PC. Bayesian extensions of the Tobit model for analyzing measures of health

status. Med Decis Making 2002; 22:152-62.

(132) Austin PC. A comparison of methods for analyzing health-related quality-of-life

measures. Value Health 2002; 5:329-37.

(133) Huang IC, Frangakis C, Atkinson MJ, Willke RJ, Leite WL, Vogel WB, Wu AW.

Addressing ceiling effects in health status measures: a comparison of techniques

applied to measures for people with HIV disease. Health Serv Res 2008; 43:327-39.

(134) Lawrence WF, Fleishman JA. Predicting EuroQoL EQ-5D preference scores from the

SF-12 Health Survey in a nationally representative sample. Med Decis Making 2004;

24:160-9.

(135) Nichol MB, Sengupta N, Globe DR. Evaluating quality-adjusted life years: estimation

of the health utility index (HUI2) from the SF-36. Med Decis Making 2001; 21:105-12.

(136) Chuang LH, Kind P. Converting the SF-12 into the EQ-5D: an empirical comparison of

methodologies. Pharmacoeconomics 2009; 27:491-505.

Page 130: Economic Evaluation of Percutaneous Coronary Intervention in

120

(137) Rivero-Arias O, Ouellet M, Gray A, Wolstenholme J, Rothwell PM, Luengo-Fernandez

R. Mapping the Modified Rankin Scale (mRS) Measurement into the Generic EuroQol

(EQ-5D) Health Outcome. Medical Decision Making 2010; 30:341-54.

(138) Jia H, Zack MM, Moriarty DG, Fryback DG. Predicting the EuroQol Group's EQ-5D

Index from CDC's "Healthy Days" in a US Sample. Medical Decision Making 10 A.D..

(139) Mortimer D, Segal L, Sturm J. Can we derive an 'exchange rate' between descriptive

and preference-based outcome measures for stroke? Results from the transfer to utility

(TTU) technique. Health Qual Life Outcomes 2009; 7:33.

(140) Cohen DJ, Taira DA, Berezin R, Cox DA, Morice MC, Stone GW, Grines CL. Cost-

effectiveness of coronary stenting in acute myocardial infarction: results from the stent

primary angioplasty in myocardial infarction (stent-PAMI) trial. Circulation 2001;

104:3039-45.

(141) Tu JV, Bowen J, Chiu M, Ko DT, Austin PC, He Y, Hopkins R, Tarride JE,

Blackhouse G, Lazzam C, Cohen EA, Goeree R. Effectiveness and safety of drug-

eluting stents in Ontario. N Engl J Med 2007; 357:1393-402.

(142) Austin PC, Mamdani MM, Stukel TA, Anderson GM, Tu JV. The use of the propensity

score for estimating treatment effects: administrative versus clinical data. Stat Med

2005; 24:1563-78.

(143) Kinnaird TD, Stabile E, Mintz GS, Lee CW, Canos DA, Gevorkian N, Pinnow EE,

Kent KM, Pichard AD, Satler LF, Weissman NJ, Lindsay J, Fuchs S. Incidence,

predictors, and prognostic implications of bleeding and blood transfusion following

percutaneous coronary interventions. Am J Cardiol 2003; 92:930-5.

(144) Sarasin FP, Gaspoz JM, Bounameaux H. Cost-effectiveness of new antiplatelet

regimens used as secondary prevention of stroke or transient ischemic attack. Arch

Intern Med 2000; 160:2773-8.

Page 131: Economic Evaluation of Percutaneous Coronary Intervention in

121

(145) King SB III, Smith SC Jr, Hirshfeld JW Jr, Jacobs AK, Morrison DA, Williams DO.

2007 focused update of the ACC/AHA/SCAI 2005 Guideline Update for Percutaneous

Coronary Intervention: a report of the American College of Cardiology/American Heart

Association Task Force on Practice Guidelines: (2007 Writing Group to Review New

Evidence and Update the 2005 ACC/AHA/SCAI Guideline Update for Percutaneous

Coronary Intervention). Circulation. 2008;117:000-000. 2007.

(146) Cohen DJ, Breall JA, Ho KK, Kuntz RE, Goldman L, Baim DS, Weinstein MC.

Evaluating the potential cost-effectiveness of stenting as a treatment for symptomatic

single-vessel coronary disease. Use of a decision-analytic model. Circulation 1994;

89:1859-74.

(147) Tenenbaum A, Motro M, Fisman EZ, Leor J, Boyko V, Mandelzweig L, Behar S.

Functional capacity impairment in patients with coronary artery disease: prevalence,

risk factors and prognosis. Cardiology 2003; 100:207-15.

(148) Cohen DJ, Breall JA, Ho KK, Kuntz RE, Goldman L, Baim DS, Weinstein MC.

Evaluating the potential cost-effectiveness of stenting as a treatment for symptomatic

single-vessel coronary disease. Use of a decision-analytic model. Circulation 1994;

89:1859-74.

(149) Folland ED, Hartigan PM, Parisi AF. Percutaneous transluminal coronary angioplasty

versus medical therapy for stable angina pectoris: outcomes for patients with double-

vessel versus single-vessel coronary artery disease in a Veterans Affairs Cooperative

randomized trial. Veterans Affairs ACME InvestigatorS. J Am Coll Cardiol 1997;

29:1505-11.

(150) Hartigan PM, Giacomini JC, Folland ED, Parisi AF. Two- to three-year follow-up of

patients with single-vessel coronary artery disease randomized to PTCA or medical

therapy (results of a VA cooperative study). Veterans Affairs Cooperative Studies

Program ACME Investigators. Angioplasty Compared to Medicine. Am J Cardiol

1998; 82:1445-50.

Page 132: Economic Evaluation of Percutaneous Coronary Intervention in

122

(151) Henderson RA, Pocock SJ, Clayton TC, Knight R, Fox KA, Julian DG, Chamberlain

DA. Seven-year outcome in the RITA-2 trial: coronary angioplasty versus medical

therapy. J Am Coll Cardiol 2003; 42:1161-70.

(152) Zeymer U, Uebis R, Vogt A, Glunz HG, Vohringer HF, Harmjanz D, Neuhaus KL.

Randomized comparison of percutaneous transluminal coronary angioplasty and

medical therapy in stable survivors of acute myocardial infarction with single vessel

disease: a study of the Arbeitsgemeinschaft Leitende Kardiologische Krankenhausarzte.

Circulation 2003; 108:1324-8.

(153) Wijeysundera HC, Nallamothu BK, Krumholz HM, Tu JV, Ko DT. Meta-analysis:

effects of percutaneous coronary intervention versus medical therapy on angina relief.

Ann Intern Med 2010; 152:370-9.

(154) Dibra A, Kastrati A, Alfonso F, Seyfarth M, Perez-Vizcayno MJ, Mehilli J, Schomig A.

Effectiveness of drug-eluting stents in patients with bare-metal in-stent restenosis:

meta-analysis of randomized trials. J Am Coll Cardiol 2007; 49:616-23.

(155) Trikalinos TA, Alsheikh-Ali AA, Tatsioni A, Nallamothu BK, Kent DM. Percutaneous

coronary interventions for non-acute coronary artery disease: a quantitative 20-year

synopsis and a network meta-analysis. Lancet 2009; 373:911-8.

(156) Vaccaro O, Eberly LE, Neaton JD, Yang L, Riccardi G, Stamler J. Impact of diabetes

and previous myocardial infarction on long-term survival: 25-year mortality follow-up

of primary screenees of the Multiple Risk Factor Intervention Trial. Arch Intern Med

2004; 164:1438-43.

(157) Laupacis A, Feeny D, Detsky AS, Tugwell PX. How attractive does a new technology

have to be to warrant adoption and utilization? Tentative guidelines for using clinical

and economic evaluations. CMAJ 1992; 146:473-81.

(158) Stettler C, Wandel S, Allemann S, Kastrati A, Morice MC, Schomig A, Pfisterer ME,

Stone GW, Leon MB, de Lezo JS, Goy JJ, Park SJ, Sabate M, Suttorp MJ, Kelbaek H,

Page 133: Economic Evaluation of Percutaneous Coronary Intervention in

123

Spaulding C, Menichelli M, Vermeersch P, Dirksen MT, Cervinka P, Petronio AS,

Nordmann AJ, Diem P, Meier B, Zwahlen M, Reichenbach S, Trelle S, Windecker S,

Juni P. Outcomes associated with drug-eluting and bare-metal stents: a collaborative

network meta-analysis. Lancet 2007; 370:937-48.

(159) Bowen J, Hopkins R, He Y, Blackhouse G, Lazzam C, Tu J, Cohen E,Tarride J-E,

Goeree R. Systematic review and cost-effectiveness analysis of drug eluting stents

compared to bare metal stents for percutaneous coronary interventions in Ontario:

interim report for the Ontario Ministry of Health and Long-Term Care. Hamilton,

Ontario, Canada: Program for Assessment of Technology in Health, McMaster

University; December 2005. Report No. HTA002-0512. Contract No. 06129. Available

at: http://www.path-hta.ca/DESreport.pdf.

(160) Mittmann N, Brown A, Seung SJ, Coyle D, Cohen E, Brophy J, Title L, Oh P.

Economic evaluation of drug eluting stents. Ottawa, Canada: Canadian Coordinating

Office for Health Technology Assessment; February 2005. Technology report No. 53.

Available at: https://

www.ccohta.ca/publications/pdf/272_drug_eluting_stents_tr_e.pdf. 2007.

(161) Brophy J, Erickson L. An economic analysis of drug eluting coronary stents: a Quebec

perspective. Montreal, Canada: Agence d'evaluation destechnologies et des modes

d'intervention en sante (AETMIS). Montreal: Canada; September 2004. AETMIS 04-

04. Available at: www.aetmis. gouv.qc.ca. 2007.

(162) Kong DF, Eisenstein EL, Sketch MH, Jr., Zidar JP, Ryan TJ, Harrington RA, Newman

MF, Smith PK, Mark DB, Califf RM. Economic impact of drug-eluting stents on

hospital systems: a disease-state model. Am Heart J 2004; 147:449-56.

(163) Bucher HC, Hengstler P, Schindler C, Guyatt GH. Percutaneous transluminal coronary

angioplasty versus medical treatment for non-acute coronary heart disease: meta-

analysis of randomised controlled trials. BMJ 2000; 321:73-7.

Page 134: Economic Evaluation of Percutaneous Coronary Intervention in

124

(164) Katritsis DG, Ioannidis JP. Percutaneous coronary intervention versus conservative

therapy in nonacute coronary artery disease: a meta-analysis. Circulation 2005;

111:2906-12.

(165) Schomig A, Mehilli J, de Waha A, Seyfarth M, Pache J, Kastrati A. A meta-analysis of

17 randomized trials of a percutaneous coronary intervention-based strategy in patients

with stable coronary artery disease. J Am Coll Cardiol 2008; 52:894-904.

(166) Hlatky MA, Boothroyd DB, Bravata DM, Boersma E, Booth J, Brooks MM, Carrie D,

Clayton TC, Danchin N, Flather M, Hamm CW, Hueb WA, Kahler J, Kelsey SF, King

SB, Kosinski AS, Lopes N, McDonald KM, Rodriguez A, Serruys P, Sigwart U,

Stables RH, Owens DK, Pocock SJ. Coronary artery bypass surgery compared with

percutaneous coronary interventions for multivessel disease: a collaborative analysis of

individual patient data from ten randomised trials. Lancet 2009; 373:1190-7.

(167) Sleeper LA, Ramanathan K, Picard MH, Lejemtel TH, White HD, Dzavik V, Tormey

D, Avis NE, Hochman JS. Functional status and quality of life after emergency

revascularization for cardiogenic shock complicating acute myocardial infarction. J Am

Coll Cardiol 2005; 46:266-73.

(168) Anderson HV, Shaw RE, Brindis RG, McKay CR, Klein LW, Krone RJ, Ho KK,

Rumsfeld JS, Smith SC, Jr., Weintraub WS. Risk-adjusted mortality analysis of

percutaneous coronary interventions by American College of Cardiology/American

Heart Association guidelines recommendations. Am J Cardiol 2007; 99:189-96.

(169) Eagle KA, Guyton RA, Davidoff R, Edwards FH, Ewy GA, Gardner TJ, Hart JC,

Herrmann HC, Hillis LD, Hutter AM, Jr., Lytle BW, Marlow RA, Nugent WC,

Orszulak TA, Antman EM, Smith SC, Jr., Alpert JS, Anderson JL, Faxon DP, Fuster V,

Gibbons RJ, Gregoratos G, Halperin JL, Hiratzka LF, Hunt SA, Jacobs AK, Ornato JP.

ACC/AHA 2004 guideline update for coronary artery bypass graft surgery: summary

article. A report of the American College of Cardiology/American Heart Association

Page 135: Economic Evaluation of Percutaneous Coronary Intervention in

125

Task Force on Practice Guidelines (Committee to Update the 1999 Guidelines for

Coronary Artery Bypass Graft Surgery). J Am Coll Cardiol 2004; 44:e213-e310.

(170) Wong SC, Minutello R, Hong MK. Neurological complications following percutaneous

coronary interventions (a report from the 2000-2001 New York State Angioplasty

Registry). Am J Cardiol 2005; 96:1248-50.

(171) Bakhai A, Hill RA, Dundar Y, Dickson R, Walley T. Percutaneous transluminal

coronary angioplasty with stents versus coronary artery bypass grafting for people with

stable angina or acute coronary syndromes. Cochrane Database Syst Rev 2005;

CD004588.

(172) Fuchs S, Stabile E, Kinnaird TD, Mintz GS, Gruberg L, Canos DA, Pinnow EE,

Kornowski R, Suddath WO, Satler LF, Pichard AD, Kent KM, Weissman NJ. Stroke

complicating percutaneous coronary interventions: incidence, predictors, and

prognostic implications. Circulation 2002; 106:86-91.

(173) Tengs TO, Wallace A. One thousand health-related quality-of-life estimates. Med Care

2000; 38:583-637.

(174) Tsevat J, Goldman L, Lamas GA, Pfeffer MA, Chapin CC, Connors KF, Lee TH.

Functional status versus utilities in survivors of myocardial infarction. Med Care 1991;

29:1153-9.

(175) Brindis RG, Douglas PS, Hendel RC, Peterson ED, Wolk MJ, Allen JM, Patel MR,

Raskin IE, Hendel RC, Bateman TM, Cerqueira MD, Gibbons RJ, Gillam LD, Gillespie

JA, Hendel RC, Iskandrian AE, Jerome SD, Krumholz HM, Messer JV, Spertus JA,

Stowers SA. ACCF/ASNC appropriateness criteria for single-photon emission

computed tomography myocardial perfusion imaging (SPECT MPI): a report of the

American College of Cardiology Foundation Quality Strategic Directions Committee

Appropriateness Criteria Working Group and the American Society of Nuclear

Cardiology endorsed by the American Heart Association. J Am Coll Cardiol 2005;

46:1587-605.

Page 136: Economic Evaluation of Percutaneous Coronary Intervention in

126

(176) Douglas PS, Khandheria B, Stainback RF, Weissman NJ, Peterson ED, Hendel RC,

Stainback RF, Blaivas M, Des Prez RD, Gillam LD, Golash T, Hiratzka LF, Kussmaul

WG, Labovitz AJ, Lindenfeld J, Masoudi FA, Mayo PH, Porembka D, Spertus JA,

Wann LS, Wiegers SE, Brindis RG, Douglas PS, Hendel RC, Patel MR, Peterson ED,

Wolk MJ, Allen JM. ACCF/ASE/ACEP/AHA/ASNC/SCAI/SCCT/SCMR 2008

appropriateness criteria for stress echocardiography: a report of the American College

of Cardiology Foundation Appropriateness Criteria Task Force, American Society of

Echocardiography, American College of Emergency Physicians, American Heart

Association, American Society of Nuclear Cardiology, Society for Cardiovascular

Angiography and Interventions, Society of Cardiovascular Computed Tomography, and

Society for Cardiovascular Magnetic Resonance endorsed by the Heart Rhythm Society

and the Society of Critical Care Medicine. J Am Coll Cardiol 2008; 51:1127-47.

(177) Douglas PS, Khandheria B, Stainback RF, Weissman NJ, Brindis RG, Patel MR,

Khandheria B, Alpert JS, Fitzgerald D, Heidenreich P, Martin ET, Messer JV, Miller

AB, Picard MH, Raggi P, Reed KD, Rumsfeld JS, Steimle AE, Tonkovic R,

Vijayaraghavan K, Weissman NJ, Yeon SB, Brindis RG, Douglas PS, Hendel RC, Patel

MR, Peterson E, Wolk MJ, Allen JM. ACCF/ASE/ACEP/ASNC/SCAI/SCCT/SCMR

2007 appropriateness criteria for transthoracic and transesophageal echocardiography: a

report of the American College of Cardiology Foundation Quality Strategic Directions

Committee Appropriateness Criteria Working Group, American Society of

Echocardiography, American College of Emergency Physicians, American Society of

Nuclear Cardiology, Society for Cardiovascular Angiography and Interventions,

Society of Cardiovascular Computed Tomography, and the Society for Cardiovascular

Magnetic Resonance endorsed by the American College of Chest Physicians and the

Society of Critical Care Medicine. J Am Coll Cardiol 2007; 50:187-204.

(178) Gibbons RJ, Miller TD, Hodge D, Urban L, Araoz PA, Pellikka P, McCully RB.

Application of appropriateness criteria to stress single-photon emission computed

tomography sestamibi studies and stress echocardiograms in an academic medical

center. J Am Coll Cardiol 2008; 51:1283-9.

Page 137: Economic Evaluation of Percutaneous Coronary Intervention in

127

(179) Hendel RC, Patel MR, Kramer CM, Poon M, Hendel RC, Carr JC, Gerstad NA, Gillam

LD, Hodgson JM, Kim RJ, Kramer CM, Lesser JR, Martin ET, Messer JV, Redberg

RF, Rubin GD, Rumsfeld JS, Taylor AJ, Weigold WG, Woodard PK, Brindis RG,

Hendel RC, Douglas PS, Peterson ED, Wolk MJ, Allen JM, Patel MR.

ACCF/ACR/SCCT/SCMR/ASNC/NASCI/SCAI/SIR 2006 appropriateness criteria for

cardiac computed tomography and cardiac magnetic resonance imaging: a report of the

American College of Cardiology Foundation Quality Strategic Directions Committee

Appropriateness Criteria Working Group, American College of Radiology, Society of

Cardiovascular Computed Tomography, Society for Cardiovascular Magnetic

Resonance, American Society of Nuclear Cardiology, North American Society for

Cardiac Imaging, Society for Cardiovascular Angiography and Interventions, and

Society of Interventional Radiology. J Am Coll Cardiol 2006; 48:1475-97.

(180) Patel MR, Spertus JA, Brindis RG, Hendel RC, Douglas PS, Peterson ED, Wolk MJ,

Allen JM, Raskin IE. ACCF proposed method for evaluating the appropriateness of

cardiovascular imaging. J Am Coll Cardiol 2005; 46:1606-13.

Page 138: Economic Evaluation of Percutaneous Coronary Intervention in

128

Appendices

Chapter 1: Appendix A Search Strategy

Search segment specific to health economics:

[Economics - MEDLINE]

exp Economics/

Economics.mp.

exp "costs and cost analysis"/

(cost$ adj analysis).mp.

exp Cost allocation/

cost allocation.mp.

exp "Cost-benefit analysis"/

(cost-benefit adj analysis).mp.

exp Cost control/

cost control.mp.

exp Cost savings/

cost savings.mp.

exp "Cost of illness"/

(cost adj2 illness).mp.

exp Cost sharing/

cost sharing.mp.

exp "deductibles and coinsurance"/

(deductibles adj coinsurance).mp.

exp Medical savings accounts/

medical savings accounts.mp.

exp Health care costs/

health care costs.mp.

exp Direct service costs/

Direct service costs.mp.

exp Drug costs/

drug costs.mp.

exp Employer health costs/

Employer health costs.mp.

exp Hospital costs/

hospital costs.mp.

exp Health expenditures/

Health expenditures.mp.

exp Capital expenditures/

Capital expenditures.mp.

exp "Value of life"/

(value adj life).mp.

exp economics, hospital/

exp economics, medical/

exp Economics, nursing/

Page 139: Economic Evaluation of Percutaneous Coronary Intervention in

129

exp Economics, pharmaceutical/

exp "fees and charges"/

exp budgets/

(low adj cost).mp.

(high adj cost).mp.

(health?care adj cost$).mp.

(fiscal or funding or financial or finance).tw.

(cost adj estimate$).mp.

(unit adj cost$).mp.

(economic$ or pharmacoeconomic$ or price$ or pricing).tw.

(cost adj variable).mp.

(cost-effect* or "cost effect*").mp.

exp Quality-Adjusted Life Years/

"quality-adjusted life years".mp.

"quality adjusted life years".mp.

qaly.mp.

(life adj years).mp.

cost utili*.mp.

(cost adj2 utili*).mp.

(cost adj2 effect*).mp.

exp models, economic/

(economic$ adj2 model$).mp.

(cost adj2 utilit*).mp.

(cost adj2 consequenc*).mp.

net benefit.mp.

(willingness adj pay).mp.

Page 140: Economic Evaluation of Percutaneous Coronary Intervention in

130

Evidence Based Medications

Log odds

ratio of Freedom

From

Angina

1

2 3 4

3.00

2.40

1.80

1.20

0.60

0.00

-0.60

-1.20

-1.80

-2.40

-3.00

Chapter 2: APPENDIX A

a: Analysis restricted to 8 non-myocardial infarction trials

b: Meta-regression, restricted to 8 non-myocardial infarction trials

Subgroup Studies PCI Medical Summary Odds Ratio

All trials 8 876/1325

(66.1%)

743/1314

(56.5%)

1.58(1.07-2.33)

p=0.04

< 1994 2 122/192

(63.5%)

72/193

(37.3%)

3.28 (1.41-7.63)

p =0.006

1995-1999 3 276/455

(60.6%)

233/461

(50.5%)

1.37 (0.73-2.59)

p = 0.33

> 2000 3 478/678

(70.5%)

438/660

(66.4%)

1.14 (0.63-2.07)

p = 0.66

Page 141: Economic Evaluation of Percutaneous Coronary Intervention in

131

Chapter 3: APPENDIX A: Multiple Imputation Analysis

Model Complete-Case Multiple

Imputation

(Markov Chain

Monte Carlo)

EQ, SAQ only

Multiple

Imputation

(Markov

Chain

Monte

Carlo)with

age/sex

Intercept 0.439 0.471 0.470

β AF 0.00101 0.000772 0.000767

β AS -0.000242 -0.0000915 -0.0000882

β DP 0.00233 0.00228 0.002250

β PL 0.00194 0.00203 -

0.000020893

β TS 0.000385 0.0000142 0.002109

R2

(validation)

0.38 0.39 0.38

Adjusted R2

(validation)

0.37 0.38 0.37

Mean

prediction error

0.088 0.089 0.091

Predicted mean

EQ-5D score

0.81 0.80 0.80

AF: Angina Frequency; AS: angina stability; DP: Disease Perception; PL: physical limitation; TS: treatment

satisfaction; R2: proportion of variance explained by the model

Page 142: Economic Evaluation of Percutaneous Coronary Intervention in

132

Chapter 3: APPENDIX B: WINBUGS CODE

Model 2: model{ for(i in 1:n){ EQ_Index[i] ~ dnorm(mu[i],tau[i]) mu[i] <- beta + beta_PL*SAQPL[i] + beta_AS*SAQAS[i] + beta_AF*SAQAF[i] + beta_TS*SAQTS[i] + beta_DP*SAQDP[i] tau[i]<- abs(alpha + alpha_PL*SAQPL[i] + alpha_AS*SAQAS[i] + alpha_AF*SAQAF[i] + alpha_TS*SAQTS[i] + alpha_DP*SAQDP[i]) error[i] <- abs(mu[i]-EQ_Index[i]) error_sq[i]<-((mu[i]-EQ_Index[i])*(mu[i]-EQ_Index[i])) total_error[i]<-((EQ_Index[i]-mean(EQ_Index[]))*(EQ_Index[i]-mean(EQ_Index[]))) } for(i in 1:v){ f[i]<-beta+ beta_PL*PL_valid[i] + beta_AS*AS_valid[i] + beta_AF*AF_valid[i] + beta_TS*TS_valid[i] + beta_DP*DP_valid[i] error_predict[i] <- abs(f[i]-eq_valid[i]) error_predict_sq[i]<-((f[i]-eq_valid[i])*(f[i]-eq_valid[i])) total_error_predict[i]<-((eq_valid[i]-mean(eq_valid[]))*(eq_valid[i]-mean(eq_valid[]))) } mean_error <- mean(error[]) sum_error <- sum(error[]) R_sq <- 1-(sum(error_sq[])/sum(total_error[])) mean_error_p<- mean(error_predict[]) sum_error_p <- sum(error_predict[]) R_sq_predict <- 1-(sum(error_predict_sq[])/sum(total_error_predict[])) beta ~ dnorm (0,0.00001) beta_PL ~ dnorm (0,0.00001) beta_AS ~ dnorm (0,0.00001) beta_AF ~ dnorm (0,0.00001) beta_TS ~ dnorm (0,0.00001) beta_DP ~ dnorm (0,0.00001) alpha~ dnorm (0,0.00001) alpha_PL ~ dnorm (0,0.00001) alpha_AS ~ dnorm (0,0.00001) alpha_AF ~ dnorm (0,0.00001) alpha_TS ~ dnorm (0,0.00001) alpha_DP ~ dnorm (0,0.00001) }

For the simple models (1, and 4), the tau[i]=tau=1/sigma2, where sigma~U(0,1)

For the model 3 and 6, tau is a linear function of age and gender. For the Tobit models (4,5, and

6), the conditional means are as specified on table 1.

Page 143: Economic Evaluation of Percutaneous Coronary Intervention in

133

Chapter 3: APPENDIX C: Subgroup Analysis for Mean Prediction in Validation dataset

Females (n=94): predicted EQ-5D score 0.78 vs observed EQ-5D score 0.77

Males (n=343): predicted EQ-5D score 0.82 vs observed EQ-5D score 0.83

≤ 65 year (n=239): predicted EQ-5D score 0.81 vs observed EQ-5D score 0.81

> 65 year (n=198): predicted EQ-5D score 0.81 vs observed EQ-5D score 0.82

Previous MI (n=199): predicted EQ-5D score 0.82 vs observed EQ-5D score 0.81

No previous MI (n=238): predicted EQ-5D score 0.81 vs observed EQ-5D score 0.82

Page 144: Economic Evaluation of Percutaneous Coronary Intervention in

134

Chapter 4: Appendix A: Medical Therapy Sub tree

Legend:

With initial medical therapy strategy, patient enters in CCS2 angina state. If medical therapy is

unsuccessful for10 consecutive months, must undergo PCI. If transition to MI-sub tree, see

Figure 4. If successful PCI, see Figure 3. Peri-procedural complications modelled include

death, stroke and major bleeding.

Page 145: Economic Evaluation of Percutaneous Coronary Intervention in

135

Chapter 4: Appendix B: PCI Sub tree

Legend:

Patient may enter the PCI sub tree initially if in the BMS, or DES strategies. In addition, patient

can enter this sub tree after PCI for failed medical therapy. If transition to MI, see Figure 4.

Page 146: Economic Evaluation of Percutaneous Coronary Intervention in

136

Chapter 4: Appendix C: MI sub tree

Legend:

All patients who suffer an MI undergo PCI. If the PCI is immediately following medical

therapy, there is an equal probability of BMS, DES. If MI is post-PCI, then DES is used. After

a successful PCI for MI, patient enters a Post-MI NYHA class 1-3 state. Patient may transition

between post-MI NYHA classes.

Page 147: Economic Evaluation of Percutaneous Coronary Intervention in

137

Chapter 4: Appendix D: CABG sub-tree

Legend:

After a total of 3 PCI for any indication, patient will undergo CABG for recurrent MI or angina.

Once in the post-CABG stable state, if subsequent angina or MI, the patient will return to

previous MI (Figure 4) and PCI (Figure 3) sub-trees respectively.

Page 148: Economic Evaluation of Percutaneous Coronary Intervention in

138

Chapter 4: Appendix E: Validation of Subgroup Analyses

Treatment Strategy

Time

point

BMS DES Medical Rx

CCN Model CCN Model Model

base-case

6 month 6.38

(5.45, 7.45)

6.34 2.98

(2.37, 3.75)

3.02 18.9

1 year 9.43

(8.30, 10.69)

9.42 5.17

(4.35, 6.15)

5.30 20.4

2 year 11.38

(10.15, 12.75)

11.38 7.60

(6.60, 8.76)

7.38 21.6

non-diabetic short lesion ( < 20 mm) and large artery (≥ 3mm)

6 month 4.97

(2.97, 8.24)

4.90 2.37

(1.19, 4.68)

2.38 18.7

1 year 6.75

(4.36, 10.38)

6.81 4.15

(2.48, 6.91)

4.23 20.1

2 year 8.20

(5.52, 12.08)

8.44 6.23

(4.11, 9.39)

6.40 21.1

non-diabetic long lesion (≥ 20 mm) and small artery (<3 mm)

6 month 9.51

(7.16, 12.56)

9.47 3.92

(2.49, 6.15)

4.00 19.05

1 year 13.23

(10.45, 16.68)

13.23 7.00

(5.00, 9.75)

7.00 20.87

2 year 14.79

(11.85, 18.38)

15.14 10.56

(8.07, 13.77)

10.18 22.12

diabetic short lesion ( < 20 mm) and large artery (≥ 3mm)

6 month 4.64

(2.44, 8.73)

4.56 2.36

(0.99, 5.57)

2.37 18.9

1 year 5.17

(2.82, 9.40)

5.32 3.79

(1.91, 7.44)

3.86 20.4

2 year 8.90

(5.63, 13.93)

8.98 7.17

(4.38, 11.61)

7.22 22.2

diabetic short lesion (< 20mm) and small artery (< 3mm)

6 month 5.97

(3.26, 10.81)

5.89 1.80

(0.59, 5.49)

1.84 18.9

1 year 8.98

(5.51, 14.46)

8.97 3.63

(1.65, 7.89)

3.65 20.54

2 year 12.12

(7.99, 18.16)

12.36 4.25

(2.05, 8.71)

6.48* 22.144

diabetic long lesion (≥ 20 mm) and large artery (≥ 3mm)

Page 149: Economic Evaluation of Percutaneous Coronary Intervention in

139

6 month 7.14

(4.21, 11.98)

7.10 4.32

(2.18, 8.45)

4.15 19

1 year 9.41

(5.96, 14.71)

9.36 6.51

(3.75, 11.18)

6.41 20.6

2 year 12.38

(8.33, 18.20)

11.24 7.08

(4.17, 11.87)

7.40 21.6

diabetic long lesion (≥ 20 mm) and small artery (<3 mm)

6 month 15.61

(11.41, 21.15)

15.65 4.18

(2.20, 7.89)

4.17 19.4

1 year 20.31

(15.53, 26.31)

20.43 8.40

(5.38, 13.01)

8.38 21.7

2 year 22.67

(17.64, 28.86)

22.67 9.83

(6.52, 14.69)

9.81 23.1

*: calibrated to 3 year data