ࡱ > " $ ! k m @ *z bjbjqq ǎ % D P
P 0 N ~! " ! ! ! 5% R % % $ | R Ϊ r ץ 2 $ @ 5% 2 2 ץ ! ! U p D D D 2 D ! ! D 2 D D L D M ! $ e" : wM Go 1 0 M X @ e? 2 @ M P P M @ gU % ) D m, , . n % $ % % ץ ץ P ( D - D ^ P ( - The Cost-Effectiveness of an Early Interventional Strategy in Non-ST-Elevation Acute Coronary Syndrome Based on the RITA 3 Trial
Technical report
Martin Henriksson1,2
David Epstein1
Stephen Palmer1
Mark Sculpher1
Tim Clayton3
Stuart Pocock3
Robert Henderson4
Martin Buxton5
Keith Fox6
1Centre for Health Economics, University of York, UK
2Center for Medical Technology Assessment, Linkping University, Sweden
3Medical Statistics Unit, London School of Hygiene and Tropical Medicine, UK
4Nottingham City Hospital HHS Trust, Nottingham UK
5Health Economics Research Group, Brunel University, Uxbridge, UK
6Centre for Cerebrovascular Science, Department of Medical and Radiological Sciences, University of Edinburgh, UK
Correspondence:
Martin Henriksson
Center for Medical Technology Assessment
Linkping University, 581 83 Linkping, Sweden
E-mail: martin.henriksson@ihs.liu.se
Tel: +46 13 224983
Fax: +46 13 224995
Abstract
Background: Evidence suggests that an early interventional strategy for patients with non-ST-elevation acute coronary syndrome (NSTE-ACS) can improve health outcomes but also increase costs when compared with a conservative strategy.
Objective: The aim of this study was to assess the cost-effectiveness of an early interventional strategy in different risk groups from a UK health-service perspective.
Design: Decision-analytic model based on randomised clinical trial data.
Main outcome measures: Costs in UK Sterling at 2003/2004 prices and quality-adjusted life years (QALYs) combined into an incremental cost-effectiveness ratio.
Methods: Data from the third Randomised Intervention Trial of unstable Angina (RITA 3) was employed to estimate rates of cardiovascular death and myocardial infarction, costs and health-related quality of life. Cost-effectiveness was estimated over patients lifetimes within the decision-analytic model.
Results: The mean incremental cost per QALY gained for an early interventional strategy was approximately 55,000, 22,000 and 12,000 for patients at low, intermediate and high risk, respectively. The early interventional strategy is approximately 1%, 35% and 95% likely to be cost-effective for patients at low, intermediate and high risk, respectively, at a threshold of 20,000 per QALY The cost-effectiveness of early intervention in low-risk patients is sensitive to assumptions about the duration of the treatment effect.
Conclusion: An early interventional strategy in patients presenting with NSTE-ACS is likely to be considered cost-effective for patients at high and intermediate risk, but this is less likely to be the case for patients at low risk.
Keywords: Non-ST-elevation acute coronary syndrome, cost effectiveness, statistical modelling, decision analysis, quality adjusted life years.
Introduction
Non-ST-elevation acute coronary syndrome (NSTE-ACS) represents a major health burden to health care systems and patients face a substantial risk of mortality and cardiovascular events. Although evidence suggests that the use of a strategy of early angiography with a view to revascularisation in the management of patients with NSTE-ACS is associated with an increased risk of myocardial infarction (MI) or death in the index hospitalisation (randomization to hospital discharge), the reduced risk subsequently implies an overall reduction in the risk of MI or death ADDIN EN.CITE Mehta2005486048615956636293232005Jun 15Routine vs selective invasive strategies in patients with acute coronary syndromes: a collaborative meta-analysis of randomized trials2908-17Department of Medicine, McMaster University, and Population Health Research Institute, Hamilton Health Sciences, Hamilton, Ontario, Canada L6K 1B8. smehta@mcmaster.caMehta, S. R.Cannon, C. P.Fox, K. A.Wallentin, L.Boden, W. E.Spacek, R.Widimsky, P.McCullough, P. A.Hunt, D.Braunwald, E.Yusuf, S.JamaAgedAngina, Unstable/*therapyCoronary AngiographyFibrinolytic Agents/therapeutic useHumansMiddle AgedMyocardial Infarction/*therapyMyocardial RevascularizationRandomized Controlled TrialsResearch Support, Non-U.S. Gov'thttp://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=15956636[1]. The third Randomised Intervention Trial of unstable Angina (RITA 3) confirmed these findings when 5-year results were reported ADDIN EN.CITE Fox200544704471615401836694892005Sep 10-165-year outcome of an interventional strategy in non-ST-elevation acute coronary syndrome: the British Heart Foundation RITA 3 randomised trial914-20Centre for Cardiovascular Science, Department of Medical and Radiological Sciences, University of Edinburgh, Edinburgh EH16 4SB, UK. k.a.a.fox@ed.ac.ukFox, K. A.Poole-Wilson, P.Clayton, T. C.Henderson, R. A.Shaw, T. R.Wheatley, D. J.Knight, R.Pocock, S. J.LancetAngina, Unstable/diagnosis/*therapyCause of DeathCoronary Angiography*ElectrocardiographyFollow-Up StudiesHumansMiddle AgedMyocardial Infarction/diagnosis/mortality/*therapyMyocardial RevascularizationResearch Support, Non-U.S. Gov'thttp://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=16154018[2]. The 5-year results showed that an early interventional strategy reduced the risk of the composite endpoint of death or MI (odds ratio 0.78, 95% CI 0.61-0.99) ADDIN EN.CITE Fox200544704471615401836694892005Sep 10-165-year outcome of an interventional strategy in non-ST-elevation acute coronary syndrome: the British Heart Foundation RITA 3 randomised trial914-20Centre for Cardiovascular Science, Department of Medical and Radiological Sciences, University of Edinburgh, Edinburgh EH16 4SB, UK. k.a.a.fox@ed.ac.ukFox, K. A.Poole-Wilson, P.Clayton, T. C.Henderson, R. A.Shaw, T. R.Wheatley, D. J.Knight, R.Pocock, S. J.LancetAngina, Unstable/diagnosis/*therapyCause of DeathCoronary Angiography*ElectrocardiographyFollow-Up StudiesHumansMiddle AgedMyocardial Infarction/diagnosis/mortality/*therapyMyocardial RevascularizationResearch Support, Non-U.S. Gov'thttp://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=16154018[2].
Furthermore, it has been shown that an early interventional strategy improves health-related quality of life at one year but also leads to increased costs when compared to a conservative strategy ADDIN EN.CITE Kim20054490449156530194522005Jan 18Health-related quality of life after interventional or conservative strategy in patients with unstable angina or non-ST-segment elevation myocardial infarction: one-year results of the third Randomized Intervention Trial of unstable Angina (RITA-3)221-8Medical Statistics Unit, London School of Hygiene & Tropical Medicine, Keppel Street, London WC1E 7HT, UK. joseph.kim@lshtm.ac.ukKim, J.Henderson, R. A.Pocock, S. J.Clayton, T.Sculpher, M. J.Fox, K. A.J Am Coll CardiolAdultAngina, Unstable/*therapyComparative StudyFollow-Up Studies*Health StatusHumansMyocardial Infarction/*therapy*Myocardial Revascularization*Quality of LifeQuestionnairesResearch Support, Non-U.S. Gov'tSeverity of Illness IndexTime FactorsTreatment Outcomehttp://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=15653019Epstein20064890489Epstein, D.Sculpher, M.Clayton, T.Henderson, R. A.Pocock, S. J.Buxton, M.Fox, K. A.2006A strategy of early angiography compared to conservative management in non-ST-elevation myocardial infarction: cost results from the third randomised intervention treatment in angina (RITA-3) trialSubmitted[3,4]. In order to establish whether an early interventional strategy should be recommended for widespread implementation, its cost-effectiveness needs to be assessed to determine whether the gain in health outcomes justifies any increased costs.
Based on data from RITA 3, rates of cardiovascular death or MI, costs and health-related quality of life were estimated within a decision analytic model. The analysis investigated heterogeneity in cost-effectiveness in patients with different risk profiles at randomization, which was of particular interest as baseline risk has been suggested as an important predictor of cardiovascular events and the effectiveness of early intervention ADDIN EN.CITE Fox200544704471615401836694892005Sep 10-165-year outcome of an interventional strategy in non-ST-elevation acute coronary syndrome: the British Heart Foundation RITA 3 randomised trial914-20Centre for Cardiovascular Science, Department of Medical and Radiological Sciences, University of Edinburgh, Edinburgh EH16 4SB, UK. k.a.a.fox@ed.ac.ukFox, K. A.Poole-Wilson, P.Clayton, T. C.Henderson, R. A.Shaw, T. R.Wheatley, D. J.Knight, R.Pocock, S. J.LancetAngina, Unstable/diagnosis/*therapyCause of DeathCoronary Angiography*ElectrocardiographyFollow-Up StudiesHumansMiddle AgedMyocardial Infarction/diagnosis/mortality/*therapyMyocardial RevascularizationResearch Support, Non-U.S. Gov'thttp://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=16154018[2]. The model also provided a tool to extrapolate the trial results to a relevant lifetime time horizon.
Methods
Overview
The decision problem under investigation concerns whether an early interventional (routine angiography followed by revascularisation if clinically indicated) or a conservative strategy (ischemia or symptom-driven angiography) should be recommended for patients presenting with NSTE-ACS. The analysis was undertaken from a UK health service perspective and costs are expressed in UK Sterling (GBP) at 2003/2004 prices. Health outcomes were estimated in terms of life expectancy and quality-adjusted life years (QALYs). Costs and health outcomes were discounted by 3.5 % per annum ADDIN EN.CITE (NICE)4120412National Institute for Clinical Excellence (NICE)[5].
A series of statistical models (referred to as equations) were estimated to determine the rates of cardiovascular death or non-fatal MI during the index hospitalisation and the remainder of the trial follow-up period. These estimates of effectiveness were then incorporated into the cost-effectiveness model which is based on a short-term decision tree and a long-term Markov structure as shown in Figure 1. The short and long-term models represent the index hospitalisation and the post-index hospitalisation respectively. Costs and QALYs were determined for the index hospitalisation and for each state in the long-term Markov structure.
Figure 1. Model structure
MI=myocardial infarction, CV=cardiovascular, CVD=cardiovascular death
In the short-term decision tree patients face a risk of the combined endpoint of cardiovascular death or MI as shown by the chance node labelled 1. A conditional probability determines if that endpoint is fatal or not, illustrated by the chance node labelled 2. Thus, three mutually exclusive outcomes were considered in the short-term tree (as indicated by the boxes in Figure 1): non-fatal myocardial infarction; cardiovascular death; and no event. These outcomes also represent health states in the long-term Markov structure described below. The probabilities of the different endpoints during the index hospitalisation are used to estimate the proportion of patients starting in each of the health states in the long-term model.
Logistic regression models (Equation 1 and 4 in Figure 1) were used to estimate the probabilities associated with each chance node. The regression models are presented in the data section below. Each outcome in the tree is associated with a cost, including the cost of treatment. The mean time of the index hospitalisation was 7.2 days in RITA 3 and for 90 % of the patients the index hospitalisation was 13 days or shorter. To simplify the modelling exercise the short-term tree was assumed to be instantaneous in time. Hence, the main purpose of the short-term tree was to distribute the analysed cohort over the starting states in the long-term Markov structure and to estimate the short-term costs associated with each treatment strategy.
In a Markov structure, hypothetical individuals reside in one of a set of mutually exclusive health states at each and every point in time ADDIN EN.CITE Sonnenberg19931080108Sonnenberg, F. A.Beck, J. R.1993Markov models in medical decision making: a practical guideMedical Decision Making134322-389406671417AdultAnticoagulants/ad [Administration & Dosage]Anticoagulants/ae [Adverse Effects]Case ReportCohort Studies*Decision Support TechniquesEmbolism/mo [Mortality]Embolism/pc [Prevention & Control]Heart Valve Prosthesis/mo [Mortality]Hemorrhage/ci [Chemically Induced]Hemorrhage/mo [Mortality]HumanKidney Transplantation/mo [Mortality]Male*Markov ChainsMonte Carlo MethodPostoperative Complications/mo [Mortality]PrognosisQuality of LifeSupport, U.S. Gov't, P.H.S.Survival RateBriggs19982260226101786641341998AprAn introduction to Markov modelling for economic evaluation397-409Health Economics Research Centre, Institute of Health Sciences, University of Oxford, England. andrew.briggs@ihs.ox.ac.ukBriggs, A.Sculpher, M.PharmacoeconomicsEconomics, Pharmaceutical/*statistics & numerical dataHuman*Markov Chains*Models, Economichttp://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=10178664[6,7]. During periods of equal length (normally referred to as Markov cycles) individuals can make a transition from one health state to another with transitions between health states being determined by transition probabilities. Each health state is associated with a cost and a health outcome. Costs and health outcomes from each Markov cycle are accumulated and summarised for the cohort at the termination of the analysis.
The Markov structure was made up of the three states: No event, Post MI and Dead, each represented by an oval in Figure 1. Note that two separate Dead states are drawn in Figure 1, representing death due to cardiovascular and non-cardiovascular causes. Yearly Markov cycles were implemented. It should be noted that if the acute phase of the disease had been included in the Markov structure, monthly cycles would probably have been required. However, since the acute phase of the disease is modelled in the short-term decision tree, yearly cycles were considered appropriate to model disease progression in the long-term.
As noted previously, the proportion starting in each state is determined by the outcome of the short-term decision tree. The majority of the patients in each treatment strategy will start the long-term model in the No event state although the proportion of individuals starting in this state will differ between the investigated strategies depending on their relative effectiveness during the index hospitalisation. Each year, individuals in the No event state face a probability of a composite endpoint of non-fatal MI or cardiovascular death (CVD) which is estimated using a Weibull proportional hazards model (Equation 2 in Figure 1). Note that the box (MI/CVD) in Figure 1 indicates that a composite event has occurred during a cycle and does not represent a formal health state since patients are then assigned to either a fatal or non-fatal state based on a separate calculation. As the patient-level data indicated a decreasing risk of a composite endpoint with respect to time from the index hospitalisation, this transition probability was made time dependent in the model. In a similar manner to the approach applied in the short-term decision tree, a conditional probability was then assigned to reflect the chance that this endpoint was fatal or not (Equation 4 in Figure 1). Although this probability was estimated from the same statistical model informing the estimate applied in the short-term decision tree, the estimates will not necessarily be the same in the short-term and longer-term models. This issue will be discussed in more detail below.
Patients having a non-fatal myocardial infarction in the model make a transition to the Post MI state. Once in the Post MI state, individuals face a risk of a second composite endpoint (Equation 3 in Figure 1). Analogous to the No event state, a decreasing risk with respect to the time elapsed from the myocardial infarction was employed for this transition. However, a different technical solution was needed in this part of the model since a myocardial infarction can occur in any cycle and therefore precludes using specific cycle numbers to model time dependence. Instead, tunnel states were employed to incorporate time dependence in this probability. Tunnel states are arranged so that they can be visited only in a fixed sequence and therefore make it possible to reduce or increase the risk of a clinical event as the time spent in a health state elapse. For instance, the year immediately after a myocardial infarction (the first year in the Post MI state in the model) is associated with the highest risk of a second composite endpoint. The following four years in the Post MI state are associated with successively lower risks of a second composite endpoint. For simplicity, only one Post MI state was drawn in Figure 2 but this state is effectively 5 states (with each state representing an additional year alive after a myocardial infarction). Provided no second composite endpoint occurs during a cycle, individuals only spend one cycle in each of the first four Post MI states before entering the fifth (last) Post MI state. Once an individual reaches the fifth Post MI state, they stay in that state for the sixth and subsequent cycles (years) after an MI if no second composite endpoint occurs. Hence, time dependency in the probability of a second composite endpoint was incorporated for the first five years after a non-fatal myocardial infarction in the model. Thereafter, the probability of a second composite endpoint is no longer dependent on the time elapsed from the first myocardial infarction. The conditional probability of a second composite endpoint being non fatal was estimated using Equation 4.
Patients suffering cardiovascular death at any time in the model move to the CV dead state. In each cycle patients also face a yearly risk of dying from non-cardiovascular causes. The death states are considered absorbing states in that once patients enter these states subsequent transitions are not allowed.
Data sources
Parameters were estimated from individual-patient data from RITA 3 with additional information on standard mortality rates and treatment effect being incorporated from standard life-tables and meta-analyses, respectively. Assumptions were employed to extrapolate costs and health outcomes beyond the 5-year period of trial in RITA 3.
The design and clinical outcomes of the RITA 3 trial have been described in detail elsewhere ADDIN EN.CITE Fox200544704471615401836694892005Sep 10-165-year outcome of an interventional strategy in non-ST-elevation acute coronary syndrome: the British Heart Foundation RITA 3 randomised trial914-20Centre for Cardiovascular Science, Department of Medical and Radiological Sciences, University of Edinburgh, Edinburgh EH16 4SB, UK. k.a.a.fox@ed.ac.ukFox, K. A.Poole-Wilson, P.Clayton, T. C.Henderson, R. A.Shaw, T. R.Wheatley, D. J.Knight, R.Pocock, S. J.LancetAngina, Unstable/diagnosis/*therapyCause of DeathCoronary Angiography*ElectrocardiographyFollow-Up StudiesHumansMiddle AgedMyocardial Infarction/diagnosis/mortality/*therapyMyocardial RevascularizationResearch Support, Non-U.S. Gov'thttp://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=16154018Fox200244804481224183136093352002Sep 7Interventional versus conservative treatment for patients with unstable angina or non-ST-elevation myocardial infarction: the British Heart Foundation RITA 3 randomised trial. Randomized Intervention Trial of unstable Angina743-51Cardiovascular Research, Department of Medical and Radiological Sciences, Royal Infirmary, Edinburgh EH3 9YW, UK. k.a.a.fox@ed.ac.ukFox, K. A.Poole-Wilson, P. A.Henderson, R. A.Clayton, T. C.Chamberlain, D. A.Shaw, T. R.Wheatley, D. J.Pocock, S. J.LancetAngina Pectoris/etiology/mortality/*therapyAtherectomy, CoronaryCardiotonic Agents/*therapeutic useCoronary Angiography*Coronary Artery BypassCoronary Disease/complications/mortality/*therapyEndpoint DeterminationFemaleGreat BritainHumansMaleMiddle AgedMyocardial Infarction/etiology/mortality/*therapyResearch Support, Non-U.S. Gov'tRisk Factorshttp://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=12241831[2,8]. Briefly, RITA 3 was a prospective, randomised multicentre trial with parallel groups, enrolling 1810 patients from 45 hospitals in England and Scotland, UK. Eligible patients had an episode of cardiac pain associated with electrocardiographic or previous arteriographic evidence of coronary artery disease, or an elevated serum cardiac marker. For all patients, the participating cardiologist had to be uncertain about the optimum treatment strategy, and continued medical treatment had to be an acceptable option. Eligible patients were randomised to either an early interventional strategy or a conservative strategy. In both treatment arms patients received optimal medical management. In the early interventional strategy, the aim was to also undertake coronary arteriography within 72 hours, with subsequent management (most notably revascularisation) guided by the angiographic findings ADDIN EN.CITE Fox200544704471615401836694892005Sep 10-165-year outcome of an interventional strategy in non-ST-elevation acute coronary syndrome: the British Heart Foundation RITA 3 randomised trial914-20Centre for Cardiovascular Science, Department of Medical and Radiological Sciences, University of Edinburgh, Edinburgh EH16 4SB, UK. k.a.a.fox@ed.ac.ukFox, K. A.Poole-Wilson, P.Clayton, T. C.Henderson, R. A.Shaw, T. R.Wheatley, D. J.Knight, R.Pocock, S. J.LancetAngina, Unstable/diagnosis/*therapyCause of DeathCoronary Angiography*ElectrocardiographyFollow-Up StudiesHumansMiddle AgedMyocardial Infarction/diagnosis/mortality/*therapyMyocardial RevascularizationResearch Support, Non-U.S. Gov'thttp://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=16154018Fox200244804481224183136093352002Sep 7Interventional versus conservative treatment for patients with unstable angina or non-ST-elevation myocardial infarction: the British Heart Foundation RITA 3 randomised trial. Randomized Intervention Trial of unstable Angina743-51Cardiovascular Research, Department of Medical and Radiological Sciences, Royal Infirmary, Edinburgh EH3 9YW, UK. k.a.a.fox@ed.ac.ukFox, K. A.Poole-Wilson, P. A.Henderson, R. A.Clayton, T. C.Chamberlain, D. A.Shaw, T. R.Wheatley, D. J.Pocock, S. J.LancetAngina Pectoris/etiology/mortality/*therapyAtherectomy, CoronaryCardiotonic Agents/*therapeutic useCoronary Angiography*Coronary Artery BypassCoronary Disease/complications/mortality/*therapyEndpoint DeterminationFemaleGreat BritainHumansMaleMiddle AgedMyocardial Infarction/etiology/mortality/*therapyResearch Support, Non-U.S. Gov'tRisk Factorshttp://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=12241831[2,8].
Analysis of effectiveness
All statistical analyses included previously identified risk factors for cardiac events measured at randomization and randomised treatment ADDIN EN.CITE Fox200544704471615401836694892005Sep 10-165-year outcome of an interventional strategy in non-ST-elevation acute coronary syndrome: the British Heart Foundation RITA 3 randomised trial914-20Centre for Cardiovascular Science, Department of Medical and Radiological Sciences, University of Edinburgh, Edinburgh EH16 4SB, UK. k.a.a.fox@ed.ac.ukFox, K. A.Poole-Wilson, P.Clayton, T. C.Henderson, R. A.Shaw, T. R.Wheatley, D. J.Knight, R.Pocock, S. J.LancetAngina, Unstable/diagnosis/*therapyCause of DeathCoronary Angiography*ElectrocardiographyFollow-Up StudiesHumansMiddle AgedMyocardial Infarction/diagnosis/mortality/*therapyMyocardial RevascularizationResearch Support, Non-U.S. Gov'thttp://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=16154018[2]. These risk factors were included as covariates in the statistical models and are shown in Table 1. A stepwise backward selection procedure was employed when estimating the statistical models. With this approach, a model including all specified covariates is estimated first and the most statistically non-significant covariate is then dropped (provided the level of significance of this covariate is higher than the pre-specified level). The model is then re-estimated in an iterative process until only statistically significant variables remain in the model. Within the statistical models, the general statistical approach was to drop non-significant covariates at the 5 % level. However, covariates of structural importance, such as the treatment covariate in Equation 1 and 2, were considered important to keep in the statistical models regardless of statistical significance.
Table 1. Covariates included in the statistical models
CovariateExplanationAgeDiscrete indicator for every 10 years over 60 years of age DiabetesIndicator of diabetes at study inclusionPrevious MIIndicator of previous MI at study inclusionSmokerIndicator of smoker at study inclusionPulseDiscrete indicator for every 5 beats per minuteST depressionIndicator of ST depression at study inclusionAnginaIndicator of angina grade 3 or 4 at study inclusionMaleIndicator of maleLeft BBBIndicator of left bundle branch block at study inclusionTreatIndicator of randomised to early interventional strategy
Logistic regression model of risk of cardiovascular death or myocardial infarction during the index hospitalisation (Equation 1)
A logistic regression model was used to estimate the risk of the combined endpoint of cardiovascular death or MI during the index hospitalisation. The index hospitalisation was defined as the time from randomization to hospital discharge
Weibull proportional hazards model of risk of cardiovascular death or myocardial infarction during the remainder of trial (Equation 2)
To estimate the risk of the combined endpoint of cardiovascular death or MI during the remainder of the trial period a time-to-event Weibull proportional hazards model was employed with the starting time set at hospital discharge ADDIN EN.CITE Collet19943751375Collet, D.1994Modelling survival data in medical researchLondonChapman & Hall[9]. In extrapolating beyond the period of trial follow-up (5 years), a conservative assumption of no continued treatment effect from the early interventional strategy was made. Different assumptions concerning the duration of the treatment effect after the 5 years of trial follow-up were investigated in alternative scenarios. Further details of the extrapolation are given in the results section below.
Weibull proportional hazards model of risk of a second composite endpoint of cardiovascular death or myocardial infarction (Equation 3)
There were insufficient patients in RITA 3 to estimate the risk of a second composite endpoint of MI or cardiovascular death following a non-fatal MI. Instead, the risks of a first composite endpoint were used, multiplied by the coefficient for the additional proportionate risk for patients who had a non-fatal MI prior to their entry into the RITA 3 trial. The rationale for this approach was that with a previous history of myocardial infarction a first event in the trial in fact represented at least a second event for these patients. No treatment effect of an early interventional strategy was included when estimating this risk which is a conservative assumption with respect to the cost-effectiveness of early intervention.
Logistic regression model of the proportion of composite endpoints being non-fatal (Equation 4)
A logistic regression model was employed to estimate the proportion of composite endpoints being non fatal. A dummy variable was used to investigate if this proportion was different between the index hospitalisation and the remainder of follow-up.
Death from non-cardiovascular causes
As the risk equations estimate the risk of dying from cardiovascular causes, patients risk of dying from non-cardiovascular causes needs to be included in the analysis. This risk was estimated using UK sex- and age-specific life-tables adjusted to exclude cardiovascular mortality ADDIN EN.CITE Department.200248710487Government Actuary Department.2002Interim life tables. Expectation of life for males in the United Kingdom, based on data for the years 2000-2002.London Available on: http://www.gad.gov.uk.Statistics20034881488National Statistics2003Review of the registrar general on deaths by cause, sex and age, in England and Wales.LondonNational Statistics[10,11]. This approach simplifies extrapolation. Furthermore, using lifetables rather than estimating non-cardiovascular causes from the trial data can be argued to better reflect the risk of non-cardiovascular causes in clinical practice due to selection criteria in the recruitment to RITA 3.
Costs
Comprehensive resource use data were collected in patients in RITA 3 up to one-year follow-up and have been described and analysed in detail elsewhere ADDIN EN.CITE Epstein20064890489Epstein, D.Sculpher, M.Clayton, T.Henderson, R. A.Pocock, S. J.Buxton, M.Fox, K. A.2006A strategy of early angiography compared to conservative management in non-ST-elevation myocardial infarction: cost results from the third randomised intervention treatment in angina (RITA-3) trialSubmitted[4]. Two standard OLS regressions were used to determine mean costs for the alternative strategies during the index hospitalisation and for the remainder of the trial. Mean costs were estimated, differentiating between management strategies, for patients with and without a composite endpoint of cardiovascular death or MI. When extrapolating beyond one year, the analysis assumed no difference between the treatment strategies in the cost of patients not experiencing the composite event.
Health-related quality of life
Health-related quality-of-life (HRQoL) data were collected in patients in RITA 3 at randomization, 4 months, 1 year, and yearly thereafter. Methods and results have been reported elsewhere ADDIN EN.CITE Kim20054490449156530194522005Jan 18Health-related quality of life after interventional or conservative strategy in patients with unstable angina or non-ST-segment elevation myocardial infarction: one-year results of the third Randomized Intervention Trial of unstable Angina (RITA-3)221-8Medical Statistics Unit, London School of Hygiene & Tropical Medicine, Keppel Street, London WC1E 7HT, UK. joseph.kim@lshtm.ac.ukKim, J.Henderson, R. A.Pocock, S. J.Clayton, T.Sculpher, M. J.Fox, K. A.J Am Coll CardiolAdultAngina, Unstable/*therapyComparative StudyFollow-Up Studies*Health StatusHumansMyocardial Infarction/*therapy*Myocardial Revascularization*Quality of LifeQuestionnairesResearch Support, Non-U.S. Gov'tSeverity of Illness IndexTime FactorsTreatment Outcomehttp://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=15653019[3]. To estimate QALYs for each treatment strategy, quality adjustment weights (utilities) are required on a scale where 0 represents death and 1 represents full health. These were obtained using the EQ-5D instrument, which was used in the trial, and employing the preferences of the UK general population ADDIN EN.CITE Brooks199626026Brooks, R.1996EuroQol: the current state of playHealth Policy37153-72Health Policy962816207Cost-Benefit AnalysisDeath*Economic Value of LifeEurope*Health Services Research/mt [Methods]Health StatusHumanModels, TheoreticalObserver Variation*Quality-Adjusted Life YearsReproducibility of ResultsDolan199735035Dolan, P.1997Modeling valuations for EuroQol health statesMedical Care35111095-108Med Care980336841Activities of Daily LivingAdultComorbidityForms and Records Control/sn [Statistics & Numerical Data]Great Britain*Health Services Research/mt [Methods]Human*Models, StatisticalPain MeasurementProbability*Quality-Adjusted Life YearsRegression AnalysisReproducibility of ResultsSelf CareSensitivity and SpecificitySeverity of Illness IndexSupport, Non-U.S. Gov'tUnconsciousness/di [Diagnosis]Unconsciousness/ep [Epidemiology][12,13]. A standard OLS regression was employed in order to estimate the mean HRQoL of patients with different risk profiles at randomization. A panel-data approach was then employed in order to estimate changes in HRQoL after randomization, differentiating between the two management strategies and whether a composite endpoint of cardiovascular death or MI had occurred. For the long-term extrapolation, no difference in HRQoL between the treatment strategies was assumed after the first year in patients not having experienced a composite endpoint. In a similar manner to the approach previously described in relation to Equation 3, this was considered to be a conservative assumption with respect to the cost-effectiveness of early intervention. The long-term decrement in HRQoL in patients who had experienced a non-fatal MI was based on the estimated HRQoL observed during the trial of patients who experienced such an event before or during the trial.
Analysis
Cost-effectiveness decision
The expected (mean) costs and health outcomes of both strategies were combined into an incremental cost-effectiveness ratio (ICER), which should be interpreted as the additional cost of generating an additional unit of health outcome ADDIN EN.CITE Karlsson19962740274Karlsson, G.Johannesson, M.1996The Decision Rules of Cost-Effectiveness AnalysisPharmacoeconomics92113-120[14]. Many health care systems will compare the ICER with a threshold value to establish whether the strategy should, in principle, be recommended for implementation ADDIN EN.CITE Claxton19992450245Claxton, K.1999The irrelevance of inference: a decision-making approach to the stochastic evaluation of health care technologiesJ Health Econ183341-64Jun105378995Bayes TheoremClinical TrialsCost-Benefit Analysis/*statistics & numerical data*Decision MakingGreat BritainHealth Care Costs/statistics & numerical dataHuman*Models, Econometric*Stochastic ProcessesTechnology Assessment, Biomedical/*economics/methodshttp://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=10537899Commonwealth Fund of New York, Harvard Center for Risk Analysis, Harvard School of Public Health, Boston, MA, USA. kclaxton@hsph.harvard.edu[15]. The National Institute for Health and Clinical Excellence in the UK uses a threshold of around 20,000 per QALY gained ADDIN EN.CITE (NICE)4120412National Institute for Clinical Excellence (NICE)[5].
Heterogeneity and uncertainty
To reflect the potential heterogeneity in cost-effectiveness in patients with different risk profiles, the estimated mean cost-effectiveness of an early interventional strategy was estimated for patients with the characteristics of each individual patient in RITA 3. These are presented as a distribution of mean cost-effectiveness across the sample of trial patients. More detailed characteristics and cost-effectiveness of illustrative patients in that distribution are then presented. In RITA 3, a multivariate predictive model for death or MI was used to define a risk score. This risk score was used to define quartiles of risk (risk groups 1 to 4), and because of the much higher event rate in the top quartile, this quartile was then further subdivided into equal-sized two-eights of risk (risk groups 4a and 4b) ADDIN EN.CITE Fox200544704471615401836694892005Sep 10-165-year outcome of an interventional strategy in non-ST-elevation acute coronary syndrome: the British Heart Foundation RITA 3 randomised trial914-20Centre for Cardiovascular Science, Department of Medical and Radiological Sciences, University of Edinburgh, Edinburgh EH16 4SB, UK. k.a.a.fox@ed.ac.ukFox, K. A.Poole-Wilson, P.Clayton, T. C.Henderson, R. A.Shaw, T. R.Wheatley, D. J.Knight, R.Pocock, S. J.LancetAngina, Unstable/diagnosis/*therapyCause of DeathCoronary Angiography*ElectrocardiographyFollow-Up StudiesHumansMiddle AgedMyocardial Infarction/diagnosis/mortality/*therapyMyocardial RevascularizationResearch Support, Non-U.S. Gov'thttp://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=16154018[2]. The patient characteristics of the patients with the median risk score in each of these five risk groups were used for the more detailed presentation of cost-effectiveness.
Uncertainty in cost-effectiveness was handled using probabilistic sensitivity analysis where inputs into the analysis are defined as probability distributions which reflect the precision with which they are estimated. Monte Carlo simulation is used to translate the precision in each input variable into a measure of uncertainty in overall cost-effectiveness ADDIN EN.CITE Claxton20022520252Claxton, K.Sculpher, M.Drummond, M.2002A rational framework for decision making by the National Institute For Clinical Excellence (NICE)Lancet3609334711Aug 31122418916http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=12241891Department of Economics and Related Studies, University of York, Heslington, York, UK[16]. Details of the probabilistic analysis and distributional assumptions are available in the appendix. Cost-effectiveness acceptability curves (CEACs) showing the probability of the early interventional strategy being cost-effective for different threshold values of cost-effectiveness are presented ADDIN EN.CITE Fenwick20043450345Fenwick, E.O'Brien, B. J.Briggs, A.2004Cost-effectiveness acceptability curves--facts, fallacies and frequently asked questionsHealth Econ135405-15May151274212Bayes TheoremCost-Benefit Analysis/*methods/statistics & numerical dataHumansQuality-Adjusted Life YearsResearch Support, Non-U.S. Gov'tUnited Stateshttp://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=15127421Centre for Evaluation of Medicines, St Joseph's Hospital, Hamilton, Canada. ealf100@york.ac.uk[17].
Alternative scenarios
Two alternative scenarios were investigated relating to the estimation of differential effectiveness. Firstly, a pooled treatment effect was estimated from all randomised clinical trials comparing early interventional and conservative strategies in NSTE-ACS ADDIN EN.CITE Spacek20025200520117921382332002FebValue of first day angiography/angioplasty in evolving Non-ST segment elevation myocardial infarction: an open multicenter randomized trial. The VINO Study230-8Cardiocenter, University Hospital Kralovske Vinohrady, 3rd Medical School of Charles University Prague, Prague, Czech Republic.Spacek, R.Widimsky, P.Straka, Z.Jiresova, E.Dvorak, J.Polasek, R.Karel, I.Jirmar, R.Lisa, L.Budesinsky, T.Malek, F.Stanka, P.Eur Heart JAged*Angioplasty, Transluminal, Percutaneous CoronaryComparative Study*Coronary AngiographyCoronary Artery Bypass*ElectrocardiographyEndpoint DeterminationExercise TestFemaleFollow-Up StudiesHumansLength of StayMaleMiddle AgedMyocardial Infarction/mortality/*radiography/*therapyResearch Support, Non-U.S. Gov'tSurvival AnalysisTreatment Outcomehttp://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=11792138Anderson1995517051775940982671995DecOne-year results of the Thrombolysis in Myocardial Infarction (TIMI) IIIB clinical trial. A randomized comparison of tissue-type plasminogen activator versus placebo and early invasive versus early conservative strategies in unstable angina and non-Q wave myocardial infarction1643-50University of Texas Health Science Center, Houston 77225, USA.Anderson, H. V.Cannon, C. P.Stone, P. H.Williams, D. O.McCabe, C. H.Knatterud, G. L.Thompson, B.Willerson, J. T.Braunwald, E.J Am Coll CardiolAdultAngina, Unstable/*therapyAngioplasty, Transluminal, Percutaneous CoronaryComparative StudyCoronary Artery BypassElectrocardiographyFemaleFollow-Up StudiesHumansMaleMiddle AgedMyocardial Infarction/diagnosis/mortality/*therapyPatient ReadmissionPlasminogen Activators/*therapeutic useRecurrenceReoperationResearch Support, U.S. Gov't, P.H.S.Risk Factors*Thrombolytic TherapyTissue Plasminogen Activator/*therapeutic usehttp://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=7594098Lagerqvist200651205121698011536895402006Sep 165-year outcomes in the FRISC-II randomised trial of an invasive versus a non-invasive strategy in non-ST-elevation acute coronary syndrome: a follow-up study998-1004Department of Cardiology and Uppsala Clinical Research Center, University Hospital, S-751 85 Uppsala, Sweden.Lagerqvist, B.Husted, S.Kontny, F.Stahle, E.Swahn, E.Wallentin, L.Lancethttp://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=16980115de Winter2005515051516162880353112005Sep 15Early invasive versus selectively invasive management for acute coronary syndromes1095-104Academisch Medisch Centrum, Amsterdam, Netherlands. r.j.dewinter@amc.uva.nlde Winter, R. J.Windhausen, F.Cornel, J. H.Dunselman, P. H.Janus, C. L.Bendermacher, P. E.Michels, H. R.Sanders, G. T.Tijssen, J. G.Verheugt, F. W.N Engl J MedAcute DiseaseAdultAgedAged, 80 and overAngina, Unstable/mortality/radiography/*therapy*Angioplasty, Transluminal, Percutaneous CoronaryComparative Study*Coronary Angiography*Coronary Artery BypassElectrocardiographyFemaleHumansMaleMiddle AgedMyocardial Infarction/mortality/prevention & control/radiography/*therapyPlatelet Aggregation Inhibitors/therapeutic useRecurrence/prevention & controlResearch Support, Non-U.S. Gov'tRiskTroponin T/bloodhttp://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=16162880Fox200544704471615401836694892005Sep 10-165-year outcome of an interventional strategy in non-ST-elevation acute coronary syndrome: the British Heart Foundation RITA 3 randomised trial914-20Centre for Cardiovascular Science, Department of Medical and Radiological Sciences, University of Edinburgh, Edinburgh EH16 4SB, UK. k.a.a.fox@ed.ac.ukFox, K. A.Poole-Wilson, P.Clayton, T. C.Henderson, R. A.Shaw, T. R.Wheatley, D. J.Knight, R.Pocock, S. J.LancetAngina, Unstable/diagnosis/*therapyCause of DeathCoronary Angiography*ElectrocardiographyFollow-Up StudiesHumansMiddle AgedMyocardial Infarction/diagnosis/mortality/*therapyMyocardial RevascularizationResearch Support, Non-U.S. Gov'thttp://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=16154018Boden199851805189632444338251998Jun 18Outcomes in patients with acute non-Q-wave myocardial infarction randomly assigned to an invasive as compared with a conservative management strategy. Veterans Affairs Non-Q-Wave Infarction Strategies in Hospital (VANQWISH) Trial Investigators1785-92Veterans Affairs Medical Center and the State University of New York Health Science Center, Syracuse 13210, USA.Boden, W. E.O'Rourke, R. A.Crawford, M. H.Blaustein, A. S.Deedwania, P. C.Zoble, R. G.Wexler, L. F.Kleiger, R. E.Pepine, C. J.Ferry, D. R.Chow, B. K.Lavori, P. W.N Engl J MedAngioplasty, Transluminal, Percutaneous CoronaryComparative Study*Coronary AngiographyCoronary Artery BypassExercise TestFemaleHumansMaleMiddle AgedMyocardial Infarction/drug therapy/mortality/surgery/*therapy*Myocardial RevascularizationProspective StudiesRadionuclide VentriculographyResearch Support, Non-U.S. Gov'tResearch Support, U.S. Gov't, Non-P.H.S.Survival AnalysisThrombolytic TherapyTreatment Outcomehttp://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=9632444McCullough1998519051997414993231998SepA prospective randomized trial of triage angiography in acute coronary syndromes ineligible for thrombolytic therapy. Results of the medicine versus angiography in thrombolytic exclusion (MATE) trial596-605Henry Ford Health System, Henry Ford Heart and Vascular Institute, Detroit, Michigan 48202, USA. pmc@mich.comMcCullough, P. A.O'Neill, W. W.Graham, M.Stomel, R. J.Rogers, F.David, S.Farhat, A.Kazlauskaite, R.Al-Zagoum, M.Grines, C. L.J Am Coll CardiolAdultAgedCause of DeathComparative Study*Coronary AngiographyCoronary Disease/mortality/*radiography/therapyFemaleHospital MortalityHumansMaleMiddle AgedMyocardial Infarction/mortality/*radiography/therapyMyocardial RevascularizationPrognosisProspective StudiesSensitivity and Specificity*Thrombolytic TherapyTreatment Outcome*Triagehttp://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=9741499Cannon2001516051611419424344252001Jun 21Comparison of early invasive and conservative strategies in patients with unstable coronary syndromes treated with the glycoprotein IIb/IIIa inhibitor tirofiban1879-87Cardiovascular Division, Brigham and Women's Hospital, Boston, MA 02115, USA. cpcannon@partners.orgCannon, C. P.Weintraub, W. S.Demopoulos, L. A.Vicari, R.Frey, M. J.Lakkis, N.Neumann, F. J.Robertson, D. H.DeLucca, P. T.DiBattiste, P. M.Gibson, C. M.Braunwald, E.N Engl J MedAgedAngina, Unstable/drug therapy/mortality/*therapy*Angioplasty, Transluminal, Percutaneous CoronaryAspirin/therapeutic useCombined Modality TherapyComparative StudyCoronary AngiographyDrug Therapy, CombinationElectrocardiographyFemaleFibrinolytic Agents/therapeutic useHeparin/therapeutic useHumansMaleMiddle AgedMyocardial Infarction/drug therapy/mortality/*therapyPlatelet Aggregation Inhibitors/*therapeutic usePlatelet Glycoprotein GPIIb-IIIa Complex/*antagonists & inhibitorsTreatment OutcomeTyrosine/*analogs & derivatives/*therapeutic usehttp://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=11419424[2,18-24]. The rationale for this analysis was that data from RITA 3 could be considered relevant to inform the baseline risk of patients in the UK. However, it could be argued, once controlling for baseline risk, that the treatment effect should be pooled from all randomised trials comparing an early interventional strategy with a conservative strategy in order to incorporate all available evidence in the cost-effectiveness model. Data for this analysis were extracted from an earlier published meta-analysis ADDIN EN.CITE Mehta2005486048615956636293232005Jun 15Routine vs selective invasive strategies in patients with acute coronary syndromes: a collaborative meta-analysis of randomized trials2908-17Department of Medicine, McMaster University, and Population Health Research Institute, Hamilton Health Sciences, Hamilton, Ontario, Canada L6K 1B8. smehta@mcmaster.caMehta, S. R.Cannon, C. P.Fox, K. A.Wallentin, L.Boden, W. E.Spacek, R.Widimsky, P.McCullough, P. A.Hunt, D.Braunwald, E.Yusuf, S.JamaAgedAngina, Unstable/*therapyCoronary AngiographyFibrinolytic Agents/therapeutic useHumansMiddle AgedMyocardial Infarction/*therapyMyocardial RevascularizationRandomized Controlled TrialsResearch Support, Non-U.S. Gov'thttp://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=15956636[1] updated with the results from the more recent ICTUS trial ADDIN EN.CITE de Winter2005515051516162880353112005Sep 15Early invasive versus selectively invasive management for acute coronary syndromes1095-104Academisch Medisch Centrum, Amsterdam, Netherlands. r.j.dewinter@amc.uva.nlde Winter, R. J.Windhausen, F.Cornel, J. H.Dunselman, P. H.Janus, C. L.Bendermacher, P. E.Michels, H. R.Sanders, G. T.Tijssen, J. G.Verheugt, F. W.N Engl J MedAcute DiseaseAdultAgedAged, 80 and overAngina, Unstable/mortality/radiography/*therapy*Angioplasty, Transluminal, Percutaneous CoronaryComparative Study*Coronary Angiography*Coronary Artery BypassElectrocardiographyFemaleHumansMaleMiddle AgedMyocardial Infarction/mortality/prevention & control/radiography/*therapyPlatelet Aggregation Inhibitors/therapeutic useRecurrence/prevention & controlResearch Support, Non-U.S. Gov'tRiskTroponin T/bloodhttp://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=16162880[21] and the long-term results of FRISC II ADDIN EN.CITE Lagerqvist200651205121698011536895402006Sep 165-year outcomes in the FRISC-II randomised trial of an invasive versus a non-invasive strategy in non-ST-elevation acute coronary syndrome: a follow-up study998-1004Department of Cardiology and Uppsala Clinical Research Center, University Hospital, S-751 85 Uppsala, Sweden.Lagerqvist, B.Husted, S.Kontny, F.Stahle, E.Swahn, E.Wallentin, L.Lancethttp://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=16980115[20] and the present RITA 3 analysis. Data were pooled employing a random-effects model, with the estimate of heterogeneity coming from the inverse-variance fixed-effect method ADDIN EN.CITE Sutton20003281328Sutton, A. J.Abrams, K. R.Jones, D. R.Sheldon, T.Song, F.2000Methods for Meta-Analysis in Medical ReseachChichesterWiley[25]. For the trials not reporting the treatment effect of the composite endpoint of myocardial infarction or cardiovascular death, the reported treatment effect of myocardial infarction and death was used as an approximation. To incorporate the estimates of the treatment effect from the meta-analyses into the cost-effectiveness model, the mean log-odds ratios and standard errors from the meta-analyses were used in the index equation and equation 1 instead of the odds/hazard ratios estimated from RITA 3 trial data. In the second alternative scenario, an interaction between treatment effect and risk at randomization was employed. In this analysis, the statistical models included an interaction between the risk score defined in RITA 3 ADDIN EN.CITE Fox200544704471615401836694892005Sep 10-165-year outcome of an interventional strategy in non-ST-elevation acute coronary syndrome: the British Heart Foundation RITA 3 randomised trial914-20Centre for Cardiovascular Science, Department of Medical and Radiological Sciences, University of Edinburgh, Edinburgh EH16 4SB, UK. k.a.a.fox@ed.ac.ukFox, K. A.Poole-Wilson, P.Clayton, T. C.Henderson, R. A.Shaw, T. R.Wheatley, D. J.Knight, R.Pocock, S. J.LancetAngina, Unstable/diagnosis/*therapyCause of DeathCoronary Angiography*ElectrocardiographyFollow-Up StudiesHumansMiddle AgedMyocardial Infarction/diagnosis/mortality/*therapyMyocardial RevascularizationResearch Support, Non-U.S. Gov'thttp://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=16154018[2] and treatment effect.
In addition, several scenario analyses were undertaken in order to assess the robustness of the assumptions made in the base-case analyses.
Statistical analyses
All statistical analyses were performed using Stata version 7 (Stata Statistical Software: Release 7.0. College Station, TX: Stata Corporation). The decision-analytic model was programmed and analysed in Microsoft Excel (Microsoft Corporation, Redmond, Washington, USA).
Results
Effectiveness
Logistic regression model of risk of cardiovascular death or myocardial infarction during the index hospitalisation (Equation 1)
Equation 1 shows that increasing age and severe angina (grade 3 or 4) were associated with an increased risk of a composite endpoint during the index hospitalisation (Table 2). Although not statistically significant, the early interventional strategy was associated with an increased risk of a composite endpoint during the index hospitalisation (odds ratio 1.520, 95% CI 0.864 - 2.675). The limited number of covariates which were significant in the equation may be due to the relatively small number of composite endpoints occurring during the index hospitalisation.
Table 2. Odds ratio of composite endpoint of cardiovascular death or myocardial infarction during the index hospitalisation
CovariateOdds ratio*95 % CITreat1.5200.864 to 2.675Age1.7311.262 to 2.374Angina1.8931.086 to 3.299Constant**0.0100.005 to 0.019*Note that an odds ratio > 1 indicates an increased risk of cardiovascular death or myocardial infarction.
**Note that the constant is the odds of an event when no covariate is updated.
While the standard errors indicate the uncertainty in the parameter estimates, correlation between the parameters is also important when propagating the uncertainty through the model in the probabilistic analysis ADDIN EN.CITE Briggs20064101410Briggs, A.Sculpher, M.Claxton, K.2006Decision modelling for health economic evaluationOxford University Press, USA[26]. Therefore, the Cholesky decomposition ADDIN EN.CITE Greene20033741374Greene, WH.2003Econometric AnalysisNew YorkPrentice Hall5th[27] of the variance-covariance matrix for the parameters of the index equation is presented in an appendix (Table A1). A detailed description of the use of the Cholesky matrix is also provided in the appendix.
As previously discussed in the methods sections, alternative scenarios for the treatment effect were investigated. The results of pooling the treatment effect of 8 trials are shown in Figure 2. The odds ratio of the pooled treatment effect from the meta-analysis was similar to the odds ratio in RITA 3 (odds ratio 1.42, 95% CI 0.97 - 2.07 in the pooled analysis compared to an odds ratio of 1.53, 95% CI 0.87 2.68 from RITA 3).
Figure 2. Forest plot of meta-analysis of treatment effect in the index hospitalisation
SHAPE \* MERGEFORMAT
The results of the statistical model including an interaction between baseline risk and treatment effect showed that a higher risk was associated with a decreasing odds ratio of a composite endpoint during the index hospitalisation (Table 3). For example, the odds ratio of a composite endpoint in the early interventional strategy compared with a conservative strategy approaches 1.76 for low risk patients (risk score approaching 0). For high risk patients (risk score approaching 1) the odds ratio tends towards 1.15.
Table 3. Odds ratio of composite endpoint of myocardial infarction or cardiovascular death during the index hospitalisation including an interaction between risk at randomization and treatment effect
CovariateLog odds ratio*95 % CITreat 0.567 -0.490 to 1.624Risk score 3.638 1.198 to 6.077Interaction treat and risk score-0.424 -3.834 to 2.985Constant*-4.593 -5.394 to -3.793*Note that the constant is the log odds of an event when no covariate is updated.
The corresponding Cholesky matrix is reported in the Appendix - Table A2.
The equations presented above estimate the odds of particular events. It should be noted that the odds of an event is the ratio of two complementary probabilities and therefore does not represent a probability required to populate the cost-effectiveness model. Hence, the estimated odds need to be transformed. To obtain the relevant probabilities (p) from equation 1, the inverse logit transformation was used ADDIN EN.CITE Greene20033741374Greene, WH.2003Econometric AnalysisNew YorkPrentice Hall5th[27] given by:
EMBED Equation.3
for the covariates, X, and the estimated coefficients on the lo g s c a l e , .
W e i b u l l p r o p o r t i o n a l h a z a r d s m o d e l o f r i s k o f c a r d i o v a s c u l a r d e a t h o r m y o c a r d i a l i n f a r c t i o n d u r i n g t h e r e m a i n d e r o f t r i a l ( E q u a t i o n 2 )
T h e f a c t t h a t t h e s h a p e p a r a m e t e r i n t h e W e i b u l l s t a t i s t i c a l m o d e l i s l e s s t h a n 1 i n d i c a t e s t h a t t h e r a t e o f the composite endpoint of cardiovascular death or MI declines as time elapses from hospital discharge (see Figure 3). This finding is consistent with other studies in this patient group ADDIN EN.CITE van Domburg19981240124van Domburg, R. T.van Miltenburg-van Zijl, A. J.Veerhoek, R. J.Simoons, M. L.Unstable angina: good long-term outcome after a complicated early courseAdultAgedAged, 80 and overAngina, Unstable/classification/*mortality/therapyFemaleHumanMaleMiddle AgeMyocardial Infarction/mortality/therapyMyocardial RevascularizationPrognosisRisk AssessmentSurvival AnalysisTreatment OutcomeJ Am Coll Cardiol19983171534-9[28].
Figure 3. Estimated hazards of cardiovascular death or myocardial infarction from hospital discharge until end of trial
Treat = 0 is conservative strategy, treat = 1 is early interventional strategy.
Note that the remaining covariates are evaluated at their mean value.
The results of the Weibull model are shown in Table 4. All risk factors but one (angina) were significant at the 5% level. However, angina was very close to significance and was kept in the Weibull model as a likelihood ratio test favoured the full model. The early interventional strategy was associated with a statistically significant lower rate of cardiovascular death or MI after the index hospitalisation (hazard ratio 0.621, 95% CI 0.464 - 0.830).
Table 4. Hazard ratio of composite endpoint of cardiovascular death or myocardial infarction from hospital discharge until end of trial
CovariateHazard ratio*95 % CIAge1.7771.499 to 2.108Diabetes1.9051.359 to 2.672Previous MI1.4711.087 to 1.990Smoker1.6511.207 to 2.258Pulse1.0621.012 to 1.114ST depression1.4231.067 to 1.913Angina1.3230.988 to 1.771Male1.3721.007 to 1.869Left BBB1.9771.169 to 3.344Treat0.6210.464 to 0.830Constant**0.0080.005 to 0.015Ancillary or shape parameter0.5790.505 to 0.664*Hazard ratio > 1 indicate an increased risk of cardiovascular death or myocardial infarction.
**Constant is the hazard at time zero.
The corresponding Cholesky matrix is reported in the appendix - Table A3
The results of pooling the treatment effect of 8 trials are shown in Figure 4. The pooled treatment effect was similar to the treatment effect observed in RITA 3 (hazard ratio 0.688, 95% CI 0.536 0.881 compared to an odds ratio of 0.621, 95% CI 0.464 0.830 from RITA 3).
Figure 4. Forest plot of meta-analysis of treatment effect in the follow-up period
The results of the statistical model including an interaction between baseline risk and treatment effect are shown in Table 5. Although not statistically significant, the interaction model showed that the positive treatment effect was more pronounced in patients with higher baseline risk. The hazard ratio of a first composite endpoint in the remainder of the trial is close to 1 when the risk score is tending towards 0 and approximately 0.21 when the risk score tends towards 1.
Table 5. Hazard ratio of composite endpoint of cardiovascular death or myocardial infarction from hospital discharge to end of trial including an interaction between risk at randomization and treatment effect
CovariateLog hazard ratio*95 % CITreat-0.035 -0.581 to 0.511Risk score4.925 3.993 to 5.857Interaction treat and risk score-1.518 -3.238 to 0.203Constant*-3.986 -4.345 to -3.626Ancillary or shape parameter-0.545 -0.682 to -0.408*Note that the constant is the hazard at time 0.
The corresponding Cholesky matrix is reported in the appendix - Table A4
The transition probabilities needed to populate the long-term Markov structure were derived from the results of the statistical models reported above. The yearly transition probability of a composite endpoint in Markov cycle t, tp(t), is given by
EMBED Equation.3 .
It should be noted that the survivor function of the Weibull distribution i s g i v e n b y
E M B E D E q u a t i o n . 3 w h e r e = X
f o r t h e c o v a r i a t e s , X , a n d t h e e s t i m a t e d a n d .
A s p r e v i o u s l y n o t e d , t h e e s t i m a t e d h a z a r d o f a c o m p o s i t e e n d p o i n t d e c l i n e s r e l a t i v e l y r a p i d l y a n d t e n d s t o w a r d s a c o n s t a n t h a z a r d a f t e r o n l y a f e w y e a r s . Therefore, a declining hazard was used for the first 5 years in the model. Thereafter, a constant hazard, with respect to time after the index hospitalisation, was implemented. This appeared reasonable given the shape of the hazard curve and that follow-up data in RITA 3 was only available for up to 5 years. The estimated hazard for year 5 was thus employed for the sixth and subsequent years in the model.
A complicating issue when estimating the hazard of a first composite endpoint was how to deal with age. The results of Equation 2 indicated a declining hazard that tended towards being constant at 5 years. However, using this constant hazard for the remainder of analysis time failed to incorporate the possible impact of age as patients get older in the model. The dummy variable for age employed in the Weibull model provided a pragmatic approach to take this into account. Every tenth year, the hazard of a composite endpoint was increased by updating the age covariate. We examined the robustness of this assumption in a separate scenario, in which the effect of age on the long-term hazard was excluded. In this scenario, the estimated hazard for the fifth year was not updated with age and a constant risk was thus employed throughout the remainder of the analysis. Furthermore, as noted above, a conservative assumption was made that the treatment effect did not last longer than the 5 years of trial follow-up with different assumptions concerning the duration of the treatment effect after trial follow-up being investigated in alternative scenarios.
The estimated probabilities of a first composite endpoint in the long-term Markov structure are shown in Figure 5 (for a 60-year-old patient setting all other covariates in Equation 2 at the mean value observed in the trial). The assumption of no continued treatment effect after 5 years is clearly seen in the figure as is the effect of employing the variable for age in order to increase the risk of a composite endpoint as patients get older in the model.
Figure 5. Probabilities of a first composite endpoint in the Markov model
Weibull proportional hazards model of risk of a second composite endpoint of cardiovascular death or myocardial infarction (Equation 3)
Equation 2 was used to estimate the risk of a second composite endpoint by updating the covariate for prior myocardial infarction (Table 4). The hazard ratio of this variable indicated that the risk of a second composite endpoint of cardiovascular death or myocardial infarction was estimated to be about 50% higher than the risk of a first composite endpoint. Using the results from the Weibull model estimated in equation 2 also imposed a logical time dependency, as patients were getting further away from their MI in the model. Technically this was achieved by employing tunnel states for the first 5 years after a non-fatal myocardial infarction. After 5 years the hazard of year 5 was employed, adjusted for age as patients get older in the model.
Logistic regression model of the proportion of composite endpoints being non-fatal (Equation 4)
All the events reported in the RITA trial (comprising a total of 244 first events and 17 second events) were included in the logistic regression model estimating the probability of a composite endpoint being non-fatal. The results showed that this probability was higher during the index hospitalisation than during the follow-up period (Table 6). This reflects the fact that patients are likely to receive prompt treatment if they experience an MI whilst in hospital. For those patients who had experienced an MI prior to the trial, the composite endpoint was more likely to be fatal. It should be noted that treatment effect was highly insignificant in this model (odds ratio 1.01, p-value = 0.95). Hence, given that a composite endpoint had occurred, the randomised treatment that patients received provided no additional explanatory power as to whether the composite endpoint was fatal and hence was excluded from the final statistical model.
Table 6. Odds ratio of a composite endpoint being non-fatal
CovariateOdds ratio*95 % CIIndex dummy3.0401.614 to 5.726Age0.6990.520 to 0.941Previous MI0.4920.286 to 0.847Constant**1.1890.720 to 1.964*Odds ratio > 1 indicates an event is more likely to be non fatal.
**Note, the constant is the odds of a composite endpoint being non fatal when no covariate is updated.
The corresponding Cholesky matrix is reported in the appendix Table A5
Similarly to equation 1, the inverse logit transformation was used to get the estimated probabilities required for the cost-effectiveness model from equation 4 ADDIN EN.CITE Greene20033741374Greene, WH.2003Econometric AnalysisNew YorkPrentice Hall5th[27].
Death from non-cardiovascular causes
The hazard of dying from non-cardiovascular causes was estimated using general UK population age-and-sex specific lifetables, adjusted to exclude cardiovascular mortality (ICD10 codes I00 to I99) ADDIN EN.CITE Department.200248710487Government Actuary Department.2002Interim life tables. Expectation of life for males in the United Kingdom, based on data for the years 2000-2002.London Available on: http://www.gad.gov.uk.Statistics20034881488National Statistics2003Review of the registrar general on deaths by cause, sex and age, in England and Wales.LondonNational Statistics[10,11]. The probabilities are shown in Table 7.
Table 7. Age and sex-specific probabilities of dying from non-cardiovascular causes
AgeMenWomenAgeMenWomen450.00170.0013740.02770.0187460.00190.0016750.02960.0194470.00220.0017760.03260.0216480.00230.0018770.03600.0239490.00270.0020780.03960.0263500.00270.0022790.04360.0290510.00290.0024800.04620.0303520.00320.0026810.05000.0334530.00340.0028820.05450.0375540.00370.0032830.06070.0418550.00410.0033840.06840.0479560.00470.0036850.07640.0523570.00520.0041860.08300.0576580.00570.0043870.08950.0641590.00640.0048880.09930.0717600.00710.0052890.10830.0798610.00770.0057900.11870.0910620.00850.0061910.12630.1010630.00930.0067920.14060.1114640.01000.0075930.15220.1232650.01200.0075940.16410.1323660.01210.0084950.19480.1601670.01350.0092960.20680.1727680.01480.0103970.22780.1840690.01670.0114980.23860.1996700.01770.0118990.24880.2128710.01990.01331000.27270.2311720.02220.0149730.02460.0167
Costs
During the index hospitalisation, the early interventional strategy was associated with a higher mean cost (mean 5,654, 95% CI 5,151 - 6,157) compared with a conservative strategy (Table 8). This additional costs was mainly due to the higher number of angiographies and revascularisations undertaken in the early interventional arm. After controlling for treatment allocation, a non-fatal myocardial infarction or death was associated with additional costs of 6,221 and 7,947, respectively, which included the costs for the administration of thrombolytic drugs, revascularisations and longer hospital stay in wards and intensive care. Covariates such as age, sex, and ST depression were also associated with higher costs during the index hospitalisation. It should be pointed out that including these covariates in the short-term tree will only influence the absolute cost level in both treatment strategies but have no effect on incremental costs.
Table 8. Estimated costs during the index hospitalisation
CovariateCoefficient95 % CIMI index6,221 4,314 to 8,128Dead index7,947 5,536 to 10,358Treat5,654 5,151 to 6,157Male1,035 516 to 1,553ST depression1,224 699 to 1,750Age 878 579 to 1,178Constant1,778 1,199 to 2,358The corresponding Cholesky matrix is reported in the appendix - Table A6
During the first year after the index hospitalisation, the early interventional strategy was associated with a lower mean cost (mean -1,106, 95% CI -1,562 to -650) compared with the conservative strategy (Table 9). This reflected the fact that more patients in the conservative strategy had further symptoms that necessitated revascularisation during this period. The results also indicated that patients had a substantially higher mean cost, irrespective of treatment allocation, if they suffered a myocardial infarction within the previous year (mean 5,467, 95% CI 3,890 - 7,044) or prior to the trial (mean 724, 95% CI 210 - 1,239).
Table 9. Estimated costs follow-up
CovariateCoefficient95 % CIMI year 1 5,467 3,890 to 7,044Treat-1,106-1,562 to -650Male 586 111 to 1,061Angina 1,034 550 to 1,518Previous MI 724 210 to 1,239Constant 2,735 2,249 to 3,220The corresponding Cholesky matrix is reported in the appendix - Table A7
Since cost data was only collected for 1-year in RITA 3, certain assumptions were necessary in order to translate the results of the cost regression into costs associated with the states in the long term Markov structure. In the cost-effectiveness model, the covariates for sex, angina, and previous MI result in the addition of a constant cost to every state in the model. These covariates are updated as patients progress through the Markov model e.g. the costs for all patients surviving 1 year after an MI have an additional cost of 724 applied for every year they survive without experiencing another event. While the treatment covariate predicts a lower cost for the No event state (1,106) in the early interventional strategy in the first year, it was unclear whether this differential would continue to exist in the long term. In the absence of longer term cost data we employed a conservative assumption towards the early interventional strategy. After 1 year we assumed that that the rate of revascularisations (the cost item contributing the most to the difference between early interventional and conservative strategy) was the same in the two strategies. Hence, the predicted cost of the early interventional strategy in the first year was applied to both strategies in the second and subsequent years for the No event state.
By updating the MI year 1 covariate a predicted cost for the first year in the Post MI state was obtained. To reflect the higher use of cardiovascular drugs, visits to GPs and hospital admissions assumed to occur during the second and subsequent years after a myocardial infarction, the previous MI covariate was updated. It was assumed that the increased cost of patients having had a previous myocardial infarction would be a good estimate of the long run increase in cost associated with being in the Post MI state. An alternative assumption was considered in a separate scenario, in which no additional costs were applied in the years following a myocardial infarction.
Health-related quality of life
At randomization, mean HRQoL (in terms of 0 to 1 utilities) were higher for males whereas diabetes, previous myocardial infarction, ST depression and angina were associated with lower HRQoL (Table 10).
Table 10. Estimated baseline utilities
CovariateCoefficient95 % CIDiabetes-0.0506-0.0915 to -0.0096Previous MI-0.0443-0.0761 to -0.0125ST depression-0.0660-0.0950 to -0.0369Angina-0.0738-0.1033 to -0.0443Male0.07270.0436 to 0.1017Constant0.69240.6636 to 0.7212Note that the constant shows the utility at randomization for a patient without any of the risk factors included in the analyses.
A negative (positive) sign indicates that the risk factor is associated with a lower (higher) utility at randomization.
The corresponding Cholesky matrix is reported in the appendix - Table A8.
Binary covariates were included to represent whether the utility measure was taken at month 4 (D4) or subsequently (D12) and an interaction term for treatment group. The model assumes, for patients who do not experience a myocardial infarction, changes in utility at one year are maintained until the end of the follow up period. This assumption is consistent with the results of Kim et al ADDIN EN.CITE Kim20054490449156530194522005Jan 18Health-related quality of life after interventional or conservative strategy in patients with unstable angina or non-ST-segment elevation myocardial infarction: one-year results of the third Randomized Intervention Trial of unstable Angina (RITA-3)221-8Medical Statistics Unit, London School of Hygiene & Tropical Medicine, Keppel Street, London WC1E 7HT, UK. joseph.kim@lshtm.ac.ukKim, J.Henderson, R. A.Pocock, S. J.Clayton, T.Sculpher, M. J.Fox, K. A.J Am Coll CardiolAdultAngina, Unstable/*therapyComparative StudyFollow-Up Studies*Health StatusHumansMyocardial Infarction/*therapy*Myocardial Revascularization*Quality of LifeQuestionnairesResearch Support, Non-U.S. Gov'tSeverity of Illness IndexTime FactorsTreatment Outcomehttp://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=15653019[3]. Binary covariates were also included to indicate whether a myocardial infarction had occurred recently (that is, within 1 year prior to the time of the follow up interview) (current MI) and a covariate indicating whether a myocardial infarction had occurred at all prior to the time of the follow-up interview, either before or during the trial (Previous MI). The model was fitted using GLS random-effects estimators (Stata command xtreg ADDIN EN.CITE STATA3940394STATA[29]). The number of patients with EQ-5D data in the follow up period was 1734 and the number of observations was 6203 indicating that each patient on average had their HRQoL measured 3.5 times.
In both treatment strategies HRQoL was improved at 4 months although an incremental gain of the early interventional strategy compared with the conservative strategy was seen (mean 0.0384, 95% CI 0.005 0.071) (Table 11). Between 4 and 12 months, HRQoL was improved further in both treatment strategies, although the incremental gain of the early interventional strategy did not reach conventional levels of statistical significance (mean 0.0177, 95% CI -0.013 0.048). A recent MI was associated with a decrement in HRQoL regardless of treatment allocation (mean -0.0353, 95% CI -0.078 0.008) and a previous MI prior to study inclusion was associated with a smaller HRQoL decrement (mean -0.0097, 95% CI -0.046 0.021).
Table 11. Estimated gain in health-related quality of life
CovariateCoefficientStandard error95 % CID4i*0.03840.01680.0054 to 0.07 1 4 D 1 2 0 . 0 3 8 3 0 . 0 0 7 6 0 . 0 2 3 4 t o 0 . 0 5 3 3 D 1 2 i * 0 . 0 1 7 7 0 . 0 1 5 4 - 0 . 0 1 2 6 t o 0 . 0 4 8 0 P r e v i o u s M I - 0 . 0 0 9 7 0 . 0 1 5 6 - 0 . 0 4 0 4 t o 0 . 0 2 0 9 C u r r e n t M I - 0 . 0 3 5 3 0 . 0 2 2 0 - 0 . 0 7 8 4 t o 0 . 0 0 7 8 C o n s t a n t 0 . 0 4 4 2 0 . 0 1 2 6 0 . 0 1 9 5 t o 0 . 0 6 8 9 B e t w e e n p a t i e n t s t a n d a r d e r r o r ( u ) 0 . 2 9 5 W i t h i n p a t i e n t s t a n d a r d e r r o r ( e ) 0 . 1 8 3 F r a c t i o n o f v a r i a n c e d u e t o u i ( ) 0 . 7 2 2 * N o t e t h a t c o e f f i c i e n t s r e p r e s e n t t h e g a i n i n u t i l i t y i n t h e e a r l y i n t e r v e n t i o n a l s t r a t e g y o v e r a n d a b o v e t h a t o f t h e c o n s e r v a t i v e s t r a t e g y
T h e c o r r e s p o n d i n g C h o l e s k y matrix is reported in the appendix - Table A9.
In a similar manner to the cost analysis, a number of assumptions were necessary in order to transfer the HRQoL estimates to quality-adjustment weights for different states in the Markov structure. Employing the baseline utility estimates and changes at 4 and 12 months for the conservative and interventional strategies, respectively, a mean utility at 4 and 12 months could be determined for each strategy. In the No event state in the Markov structure we used the mean of these two values for the first year and the absolute value at 12 months for the second and subsequent years. If no myocardial infarction occurred, a conservative assumption of no difference in HRQoL between the two strategies after 12 months was employed. Applying the coefficient for the current MI covariate to the baseline utilities provided an estimate of the utility to be attached to the first year in the Post MI state. The previous MI covariate was applied to the baseline utility to provide a utility for the second and subsequent years in the Post MI state.
Cost-effectiveness
Base-case analysis
The predicted mean cost-effectiveness for patients with the characteristics of each individual patient in RITA 3 is shown in Figure 6. Substantial heterogeneity in the mean cost-effectiveness between patients with different characteristics can be observed. Using a threshold of around 20,000 per QALY, it is clear from Figure 6 that an early interventional strategy is cost-effective for more patients in the higher risk groups, but there is still substantial variation in cost-effectiveness within these risk groups. The cost-effectiveness results of illustrative patients representing each risk group are shown in Table 12 and the results of the probabilistic assessment are summarized in the cost-effectiveness acceptability curves in Figure 7. The results show that an early interventional strategy has a favourable ICER and a high probability of being cost-effective for the illustrative patients in the fourth quartile of risk (both lower and upper risk halves of the quartile), with the opposite being the case for the patient representing the first quartile of risk. For the patients representing the second and third quartiles of risk, the ICERs are close to conventional typical threshold values suggesting the cost-effectiveness of early intervention in these patients is finely balanced.
Figure 6. Cost-effectiveness based on estimated mean costs and QALYs, with and without early intervention, for patients in RITA 3
ICER = incremental cost-effectiveness ratio.
Risk groups based on predicted risk of death or MI as defined in RITA 3 ADDIN EN.CITE Fox200544704471615401836694892005Sep 10-165-year outcome of an interventional strategy in non-ST-elevation acute coronary syndrome: the British Heart Foundation RITA 3 randomised trial914-20Centre for Cardiovascular Science, Department of Medical and Radiological Sciences, University of Edinburgh, Edinburgh EH16 4SB, UK. k.a.a.fox@ed.ac.ukFox, K. A.Poole-Wilson, P.Clayton, T. C.Henderson, R. A.Shaw, T. R.Wheatley, D. J.Knight, R.Pocock, S. J.LancetAngina, Unstable/diagnosis/*therapyCause of DeathCoronary Angiography*ElectrocardiographyFollow-Up StudiesHumansMiddle AgedMyocardial Infarction/diagnosis/mortality/*therapyMyocardial RevascularizationResearch Support, Non-U.S. Gov'thttp://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=16154018[2].
Table 12. Cost-effectiveness results by patient risk profile
Risk group 1Risk group 2Risk group 3Risk group 4aRisk group 4bAge 4552526166DiabetesNoNoNoNoYesPrevious myocardial infarctionNoNoYesYesYesSmokerNoYesNoYesNoPulse (beats per minute)7282828797ST depressionNoNoYesYesYesAngina (grade 3 or 4)YesNoYesNoNoMaleFemaleMaleMaleMaleMaleLeft bundle branch blockNoNoNoNoNoIncremental cost ()4,8854,8986,0456,5386,530Incremental QALY0.09090.21340.28340.54680.5122ICER ()53,76022,94921,32511,95712,750Illustrative patients based on predicted risk of death or MI as defined in RITA 3 represent each risk group ADDIN EN.CITE Fox200544704471615401836694892005Sep 10-165-year outcome of an interventional strategy in non-ST-elevation acute coronary syndrome: the British Heart Foundation RITA 3 randomised trial914-20Centre for Cardiovascular Science, Department of Medical and Radiological Sciences, University of Edinburgh, Edinburgh EH16 4SB, UK. k.a.a.fox@ed.ac.ukFox, K. A.Poole-Wilson, P.Clayton, T. C.Henderson, R. A.Shaw, T. R.Wheatley, D. J.Knight, R.Pocock, S. J.LancetAngina, Unstable/diagnosis/*therapyCause of DeathCoronary Angiography*ElectrocardiographyFollow-Up StudiesHumansMiddle AgedMyocardial Infarction/diagnosis/mortality/*therapyMyocardial RevascularizationResearch Support, Non-U.S. Gov'thttp://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=16154018[2].
QALY = quality-adjusted life year, ICER = incremental cost-effectiveness ratio
Figure 7. Cost-effectiveness acceptability curves for the early interventional strategy
Illustrative patients based on predicted risk of death or MI as defined in RITA 3 represent each risk group ADDIN EN.CITE Fox200544704471615401836694892005Sep 10-165-year outcome of an interventional strategy in non-ST-elevation acute coronary syndrome: the British Heart Foundation RITA 3 randomised trial914-20Centre for Cardiovascular Science, Department of Medical and Radiological Sciences, University of Edinburgh, Edinburgh EH16 4SB, UK. k.a.a.fox@ed.ac.ukFox, K. A.Poole-Wilson, P.Clayton, T. C.Henderson, R. A.Shaw, T. R.Wheatley, D. J.Knight, R.Pocock, S. J.LancetAngina, Unstable/diagnosis/*therapyCause of DeathCoronary Angiography*ElectrocardiographyFollow-Up StudiesHumansMiddle AgedMyocardial Infarction/diagnosis/mortality/*therapyMyocardial RevascularizationResearch Support, Non-U.S. Gov'thttp://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=16154018[2].
Alternative scenarios
The estimated effectiveness and cost-effectiveness for the alternative scenarios are shown in Tables 13 and 14 for the same patients representing the different risk groups in the base-case scenario. The pooled treatment effects from the meta-analyses of 8 trials (both in the index and follow-up period) were similar to the treatment effect observed in RITA 3. Hence, the estimated cost-effectiveness using this alternative scenario is similar to that observed when using the treatment effect estimated in RITA 3 (Table 13).
As noted above, the results of the interaction model showed that higher risk was associated with a decreasing odds ratio of a composite endpoint during the index hospitalisation and a more pronounced treatment effect in the follow-up period. Consequently, the cost-effectiveness in patients at high risk was somewhat improved compared to the base-case scenario of a common treatment effect. Conversely, for patients at low risk, cost-effectiveness was much less favourable comparing with the base-case scenario (Table 14).
Table 13. Cost-effectiveness results using a pooled treatment effect from 8 trials
Risk group 1Risk group 2Risk group 3Risk group 4aRisk group 4bOdds ratio index hospitalisation with early intervention1.421.421.421.421.42Hazard ratio in follow-up period with early intervention0.690.690.690.690.69Incremental cost ()4,8194,8525,7886,1636,129Incremental QALY0.08240.18470.23970.45170.4178ICER ()58,49026,26524,14313,64614,673Illustrative patients based on predicted risk of death or MI as defined in RITA 3 represent each risk group ADDIN EN.CITE Fox200544704471615401836694892005Sep 10-165-year outcome of an interventional strategy in non-ST-elevation acute coronary syndrome: the British Heart Foundation RITA 3 randomised trial914-20Centre for Cardiovascular Science, Department of Medical and Radiological Sciences, University of Edinburgh, Edinburgh EH16 4SB, UK. k.a.a.fox@ed.ac.ukFox, K. A.Poole-Wilson, P.Clayton, T. C.Henderson, R. A.Shaw, T. R.Wheatley, D. J.Knight, R.Pocock, S. J.LancetAngina, Unstable/diagnosis/*therapyCause of DeathCoronary Angiography*ElectrocardiographyFollow-Up StudiesHumansMiddle AgedMyocardial Infarction/diagnosis/mortality/*therapyMyocardial RevascularizationResearch Support, Non-U.S. Gov'thttp://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=16154018[2].
QALY = quality-adjusted life year, ICER = incremental cost-effectiveness ratio
Table 14. Cost-effectiveness results when treatment effect is estimated using an
interaction between risk at randomization in RITA 3 and treatment effect
Risk
group 1Risk group 2Risk group 3Risk group 4aRisk group 4bOdds ratio index hospitalisation with early intervention1.711.671.671.561.47Hazard ratio in follow-up period with early intervention0.860.800.720.620.50Incremental cost ()4,7464,7745,5746,5527,214Incremental QALY-0.01850.09520.18760.55070.6886ICER ()Dominated50,13129,71111,89810,476Illustrative patients based on predicted risk of death or MI as defined in RITA 3 represent each risk group ADDIN EN.CITE Fox200544704471615401836694892005Sep 10-165-year outcome of an interventional strategy in non-ST-elevation acute coronary syndrome: the British Heart Foundation RITA 3 randomised trial914-20Centre for Cardiovascular Science, Department of Medical and Radiological Sciences, University of Edinburgh, Edinburgh EH16 4SB, UK. k.a.a.fox@ed.ac.ukFox, K. A.Poole-Wilson, P.Clayton, T. C.Henderson, R. A.Shaw, T. R.Wheatley, D. J.Knight, R.Pocock, S. J.LancetAngina, Unstable/diagnosis/*therapyCause of DeathCoronary Angiography*ElectrocardiographyFollow-Up StudiesHumansMiddle AgedMyocardial Infarction/diagnosis/mortality/*therapyMyocardial RevascularizationResearch Support, Non-U.S. Gov'thttp://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=16154018[2].
QALY = quality-adjusted life year, ICER = incremental cost-effectiveness ratio
The cost-effectiveness acceptability curves for these alternative scenarios are shown in Figures 8 and 9.
Figure 8. Cost-effectiveness acceptability curves for the early interventional strategy
using a pooled treatment effect from 8 trials
Illustrative patients based on predicted risk of death or MI as defined in RITA 3 represent each risk group ADDIN EN.CITE Fox200544704471615401836694892005Sep 10-165-year outcome of an interventional strategy in non-ST-elevation acute coronary syndrome: the British Heart Foundation RITA 3 randomised trial914-20Centre for Cardiovascular Science, Department of Medical and Radiological Sciences, University of Edinburgh, Edinburgh EH16 4SB, UK. k.a.a.fox@ed.ac.ukFox, K. A.Poole-Wilson, P.Clayton, T. C.Henderson, R. A.Shaw, T. R.Wheatley, D. J.Knight, R.Pocock, S. J.LancetAngina, Unstable/diagnosis/*therapyCause of DeathCoronary Angiography*ElectrocardiographyFollow-Up StudiesHumansMiddle AgedMyocardial Infarction/diagnosis/mortality/*therapyMyocardial RevascularizationResearch Support, Non-U.S. Gov'thttp://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=16154018[2].
QALY = quality-adjusted life year.
Figure 9. Cost-effectiveness acceptability curves for the early interventional strategy
when treatment effect is estimated using an interaction between risk at
randomization in RITA 3 and treatment effect
Illustrative patients based on predicted risk of death or MI as defined in RITA 3 represent each risk group ADDIN EN.CITE Fox200544704471615401836694892005Sep 10-165-year outcome of an interventional strategy in non-ST-elevation acute coronary syndrome: the British Heart Foundation RITA 3 randomised trial914-20Centre for Cardiovascular Science, Department of Medical and Radiological Sciences, University of Edinburgh, Edinburgh EH16 4SB, UK. k.a.a.fox@ed.ac.ukFox, K. A.Poole-Wilson, P.Clayton, T. C.Henderson, R. A.Shaw, T. R.Wheatley, D. J.Knight, R.Pocock, S. J.LancetAngina, Unstable/diagnosis/*therapyCause of DeathCoronary Angiography*ElectrocardiographyFollow-Up StudiesHumansMiddle AgedMyocardial Infarction/diagnosis/mortality/*therapyMyocardial RevascularizationResearch Support, Non-U.S. Gov'thttp://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=16154018[2].
QALY = quality-adjusted life year.
The results of running different scenarios testing the importance of some of the assumptions employed for the base-case analysis of the cost-effectiveness model are shown in Table 15 and 16. The base-case analysis assumed that the treatment effect of the early interventional strategy did not continue after 5 years. To examine the robustness of the results to this assumption, three different scenarios were investigated regarding the duration of the treatment effect: a continued treatment effect for an additional 5 years (10 years in total); for an additional 10 years (15 years in total) and patients remaining lifetime. As anticipated, extending the duration of the treatment effect after the 5 years of trial follow-up had a favourable impact on the cost-effectiveness results for all risk groups (Table 15). As shown in Table 15, the importance of treatment duration for the cost-effectiveness results varies according to patients risk and the assumption applied to the relative treatment effect (constant or interaction term). Applying a threshold of 20,000 per QALY, the early interventional strategy was cost-effective for the patient representing risk group 3 when the treatment duration was assumed to continue for 10 years in total, regardless of whether a constant or interaction model was applied to the treatment effect. For the patient representing risk group 2, early intervention was cost-effective assuming the treatment effect was maintained for 10 years assuming a constant treatment effect. However, for the interaction model, early intervention was only cost-effective when the treatment effect was continued for a lifetime. For the illustrative patient in risk group 1, early intervention was considered cost-effective when a lifetime treatment effect was assumed (constant treatment effect model); in the interaction model, early intervention did not appear cost-effective regardless of the duration applied to the treatment effect. The patients representing the highest risk groups (risk groups 4a and 4b) remained cost-effective in all the scenarios considered.
Table 15. Results when using different assumptions for the duration of treatment effect after the 5-year trial period
Treatment effect scenarioRisk groupAssumption of duration of treatment effect Base-case*10 years15 yearslifetimeConstant RITA 3 treatment effect153,76034,90127,94913,920222,94915,41011,6527,850321,32515,75413,15910,4734a11,9579,6318,4467,6004b12,7509,7078,9048,270Interaction between treatment effect and risk at randomisation1Dominated187,947121,04445,130250,13128,16321,55314,354329,71119,68116,21812,7814a11,8989,4508,3347,6004b10,4767,9347,3486,906Results are expressed as cost per QALY of an interventional strategy compared with a conservative strategy. Illustrative patients based on predicted risk of death or MI as defined in RITA 3 represent each risk group ADDIN EN.CITE Fox200544704471615401836694892005Sep 10-165-year outcome of an interventional strategy in non-ST-elevation acute coronary syndrome: the British Heart Foundation RITA 3 randomised trial914-20Centre for Cardiovascular Science, Department of Medical and Radiological Sciences, University of Edinburgh, Edinburgh EH16 4SB, UK. k.a.a.fox@ed.ac.ukFox, K. A.Poole-Wilson, P.Clayton, T. C.Henderson, R. A.Shaw, T. R.Wheatley, D. J.Knight, R.Pocock, S. J.LancetAngina, Unstable/diagnosis/*therapyCause of DeathCoronary Angiography*ElectrocardiographyFollow-Up StudiesHumansMiddle AgedMyocardial Infarction/diagnosis/mortality/*therapyMyocardial RevascularizationResearch Support, Non-U.S. Gov'thttp://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=16154018[2].
*Assumes no continued treatment effect after 5 years of trial follow-up.
Testing other assumptions than that of duration of treatment effect showed that the results were robust to a series of alternative assumptions, with the estimated ICERs only being marginally altered (Table 16).
Table 16. Result of sensitivity scenarios
Scenario analysedRisk groupICER*Base-case analysis153,760222,949321,3254a11,9574b12,750No additional cost of being in the post MI state second and subsequent years after an MI149,972223,796322,1184a12,3534b13,150No age effect in the long-term probability of a composite endpoint estimated by equation 2148,452220,393318,9994a10,8334b10,986*Calculated as cost-per gained QALY of an early interventional strategy compared with a conservative strategy. Illustrative patients based on predicted risk of death or MI as defined in RITA 3 represent each risk group ADDIN EN.CITE Fox200544704471615401836694892005Sep 10-165-year outcome of an interventional strategy in non-ST-elevation acute coronary syndrome: the British Heart Foundation RITA 3 randomised trial914-20Centre for Cardiovascular Science, Department of Medical and Radiological Sciences, University of Edinburgh, Edinburgh EH16 4SB, UK. k.a.a.fox@ed.ac.ukFox, K. A.Poole-Wilson, P.Clayton, T. C.Henderson, R. A.Shaw, T. R.Wheatley, D. J.Knight, R.Pocock, S. J.LancetAngina, Unstable/diagnosis/*therapyCause of DeathCoronary Angiography*ElectrocardiographyFollow-Up StudiesHumansMiddle AgedMyocardial Infarction/diagnosis/mortality/*therapyMyocardial RevascularizationResearch Support, Non-U.S. Gov'thttp://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=16154018[2].
Model validation
Several analyses were performed to assess the validity of the model. A good indicator of overall validity is to investigate the predicted undiscounted life expectancy from the model. The results of this analysis are shown in Figure 10. Predicted life expectancy for the illustrative patient characteristics with different starting ages of the cohort is shown in the figure together with the life expectancy of the general population in the UK. As expected, predicted life expectancy decreases when the analysed risk groups have a higher risk at baseline. It can be seen in Figure 10 that the estimated life expectancy of patients at low risk (first quartile) is similar to the general population which was also expected.
Figure 10. Predicted life expectancy for different risk profiles and the general UK population
Illustrative patients based on predicted risk of death or MI as defined in RITA 3 represent each risk group ADDIN EN.CITE Fox200544704471615401836694892005Sep 10-165-year outcome of an interventional strategy in non-ST-elevation acute coronary syndrome: the British Heart Foundation RITA 3 randomised trial914-20Centre for Cardiovascular Science, Department of Medical and Radiological Sciences, University of Edinburgh, Edinburgh EH16 4SB, UK. k.a.a.fox@ed.ac.ukFox, K. A.Poole-Wilson, P.Clayton, T. C.Henderson, R. A.Shaw, T. R.Wheatley, D. J.Knight, R.Pocock, S. J.LancetAngina, Unstable/diagnosis/*therapyCause of DeathCoronary Angiography*ElectrocardiographyFollow-Up StudiesHumansMiddle AgedMyocardial Infarction/diagnosis/mortality/*therapyMyocardial RevascularizationResearch Support, Non-U.S. Gov'thttp://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=16154018[2].
In another check of model validation, the predicted number of events in the model after 5 years was compared with the number of events reported in RITA 3 (where results from 5 years follow-up were reported). In this analysis, the treatment effect observed in RITA 3 was employed and the covariates included in the risk equations were set at their mean value in the trial. Under these circumstances, the predicted number of events and odds ratios from the model is expected to be relatively close to those reported in RITA 3. The results are shown in Table 17. The model slightly underestimates the number of cardiovascular deaths but appear to predict the events in the RITA trial with reasonable precision. The odds ratios from the model are very similar to those observed in RITA 3. The marginally higher odds ratios observed in the model for some outcomes indicate that the results of the model do not appear to bias the results in favour of the early interventional strategy.
Table 17. Predicted number of events and odds ratios from the model compared with the results reported in RITA 3
InterventionalConservativeOdds ratioDeath/MIRITA 3 trial1421780.78Model1371710.79CVD/MIRITA 3 trial1051390.74Model991330.74DeathRITA 3 trial1021320.76Model911140.79CVDRITA 3 trial62900.68Model53760.70CVD = cardiovascular death, MI = myocardial infarction.
Figure 11 shows the estimated cumulative risk of a composite endpoint of death or MI in different risk groups. Again, the illustrative patients in each quartile of risk defined in RITA 3 are used in the analysis. The model appears to predict this outcome with reasonable precision in the different risk groups as the cumulative risk curves correspond well with those reported in RITA 3 (figure 5 in the clinical report ADDIN EN.CITE Fox200544704471615401836694892005Sep 10-165-year outcome of an interventional strategy in non-ST-elevation acute coronary syndrome: the British Heart Foundation RITA 3 randomised trial914-20Centre for Cardiovascular Science, Department of Medical and Radiological Sciences, University of Edinburgh, Edinburgh EH16 4SB, UK. k.a.a.fox@ed.ac.ukFox, K. A.Poole-Wilson, P.Clayton, T. C.Henderson, R. A.Shaw, T. R.Wheatley, D. J.Knight, R.Pocock, S. J.LancetAngina, Unstable/diagnosis/*therapyCause of DeathCoronary Angiography*ElectrocardiographyFollow-Up StudiesHumansMiddle AgedMyocardial Infarction/diagnosis/mortality/*therapyMyocardial RevascularizationResearch Support, Non-U.S. Gov'thttp://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=16154018[2]).
Figure 11. Cumulative risk of the composite endpoint of death or myocardial infarction in different risk groups
Illustrative patients based on predicted risk of death or MI as defined in RITA 3 represent each risk group ADDIN EN.CITE Fox200544704471615401836694892005Sep 10-165-year outcome of an interventional strategy in non-ST-elevation acute coronary syndrome: the British Heart Foundation RITA 3 randomised trial914-20Centre for Cardiovascular Science, Department of Medical and Radiological Sciences, University of Edinburgh, Edinburgh EH16 4SB, UK. k.a.a.fox@ed.ac.ukFox, K. A.Poole-Wilson, P.Clayton, T. C.Henderson, R. A.Shaw, T. R.Wheatley, D. J.Knight, R.Pocock, S. J.LancetAngina, Unstable/diagnosis/*therapyCause of DeathCoronary Angiography*ElectrocardiographyFollow-Up StudiesHumansMiddle AgedMyocardial Infarction/diagnosis/mortality/*therapyMyocardial RevascularizationResearch Support, Non-U.S. Gov'thttp://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=16154018[2].
Several tests were also performed to check for errors in programming and incorporation of data into the model. In the simplest of these tests no treatment effect was included in the model, i.e. the odds and hazard ratios were set to 1. This yielded the expected results of no difference in life-expectancy between the treatment strategies. Further excluding any differences in costs and health-related quality of life yielded the expected results of no difference at all in costs and health outcome in a lifetime perspective between the two treatment strategies.
Discussion
The analysis shows that, in patients presenting with NSTE-ACS at high risk of further cardiac events, an early interventional strategy is associated with a gain in QALYs at an additional which is likely to be considered acceptable cost when compared with a conservative strategy. However, for patients at low risk, an early interventional strategy is associated with a high cost per QALY gained. For patients at intermediate risk, the cost per QALY gained is close to generally accepted thresholds, so decisions about cost-effectiveness are likely to be finely balanced.
An important feature of the methods used in the analysis is that, as part of the scenario analysis, all available trial evidence was synthesised. This allowed evidence from 7 trials (additional to RITA-3) to be reflected in the cost-effectiveness estimates. This re-estimation of the treatment effects had little effect on the estimated cost-effectiveness, but the analysis more accurately reflects the uncertainty in the treatment effect parameters as all randomised evidence is taken into account. The usefulness of pooling the treatment effects of broad treatment strategies like the early interventional strategy investigated in this analysis is a matter of discussion. Major improvements in cardiac care, such as the introduction of stenting in percutaneous coronary interventions and Glycoprotein IIb/IIIa inhibitors, suggest that some of the more recent trials more accurately represent current clinical practice, and that more recent trials may be more relevant when estimating the treatment effect for the model. An alternative view is that pooling all relevant trials in a random-effect meta-analysis more fully takes into account the heterogeneity inherent in such a broadly defined treatment strategy as an early interventional strategy. Whatever view taken on this issue, small differences in treatment effect were observed when comparing the pooled treatment effect to that observed in RITA 3.
A key assumption in the base-case analysis is that the relative reduction in the risk of events conferred by an early intervention strategy is common across patients with different risk profiles. This is an assumption made in a number of cost-effectiveness studies in the cardiac field ADDIN EN.CITE HeartProtection20065220522HeartProtection2006Lifetime cost effectiveness of simvastatin in a range of risk groups and age groups derived from a randomised trial of 20 536 peopleBMJBriggs20065210521171352232006Nov 29The Cost-effectiveness of perindopril in reducing cardiovascular events in patients with stable coronary artery disease using data from the EUROPA StudyUniversity of Glasgow, United Kingdom.Briggs, A.Mihaylova, B.Sculpher, M.Hall, A.Wolstenholme, J.Simoons, M.Deckers, J.Roberto, F.Remme, W. J.Bertrand, M.Fox, K.Hearthttp://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=17135223[30,31]. Despite this assumption, the cost-effectiveness of early intervention varies according to patients risk profile as the latter determines their absolute benefit from early intervention which drives cost-effectiveness. The appropriateness of the assumption of a common treatment effect across risk profiles has been questioned in the clinical analysis of RITA 3 ADDIN EN.CITE Fox200544704471615401836694892005Sep 10-165-year outcome of an interventional strategy in non-ST-elevation acute coronary syndrome: the British Heart Foundation RITA 3 randomised trial914-20Centre for Cardiovascular Science, Department of Medical and Radiological Sciences, University of Edinburgh, Edinburgh EH16 4SB, UK. k.a.a.fox@ed.ac.ukFox, K. A.Poole-Wilson, P.Clayton, T. C.Henderson, R. A.Shaw, T. R.Wheatley, D. J.Knight, R.Pocock, S. J.LancetAngina, Unstable/diagnosis/*therapyCause of DeathCoronary Angiography*ElectrocardiographyFollow-Up StudiesHumansMiddle AgedMyocardial Infarction/diagnosis/mortality/*therapyMyocardial RevascularizationResearch Support, Non-U.S. Gov'thttp://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=16154018[2]. This showed a clear beneficial treatment effect on the composite endpoint of death or MI of an early interventional strategy for patients in the upper risk half in the fourth quartile of risk (odds ratio 0.44, 95% CI 0.25 0.76). In the lower risk half in the fourth quartile of risk and in the third quartile of risk a positive but not statistically significant treatment effect was observed whereas in the first and second quartile of risk the odds ratios were very close to 1. As part of the scenario analysis, the interactions between patients baseline risk and the effectiveness of early intervention were incorporated into the cost-effectiveness model. This confirmed the findings that an early interventional strategy is likely to be considered cost-effective in patients at high risk but not so in patients at low risk. Furthermore, those at intermediate risk appear less cost-effective to manage with early intervention than under the assumption of a common treatment effect.
It should be noted that the five illustrative patients for whom cost-effectiveness is reported in Table 4 and 5 only represent an estimate of the mean cost-effectiveness of these risk groups. Within each risk group substantial variation can be observed as shown in Figure 6. This is most prominent for patients at intermediate risk (second and third quartile of risk). Therefore the results presented here by risk group should be seen as indicative. In deciding in which types of patients early intervention should be used, health care decision makers will need to consider specific combinations of characteristics that affect patients risks, or see baseline risk as a continuous measure rather than as categorical.
A number of previous studies have reported on the cost-effectiveness of an early interventional strategy. The cost-effectiveness analysis conducted alongside the TACTICS trial reported 6-months within-trial results and lifetime extrapolations ADDIN EN.CITE Mahoney2002492049212377083288152002Oct 16Cost and cost-effectiveness of an early invasive vs conservative strategy for the treatment of unstable angina and non-ST-segment elevation myocardial infarction1851-8Emory Center for Outcomes Research, Division of Cardiology, Department of Medicine, Emory University School of Medicine, 1256 Briarcliff Rd, Suite 1N, Atlanta, GA 30306, USA. emahone@emory.eduMahoney, E. M.Jurkovitz, C. T.Chu, H.Becker, E. R.Culler, S.Kosinski, A. S.Robertson, D. H.Alexander, C.Nag, S.Cook, J. R.Demopoulos, L. A.DiBattiste, P. M.Cannon, C. P.Weintraub, W. S.JamaAgedAngina, Unstable/*economics/*therapyComparative StudyCost-Benefit AnalysisFemaleFollow-Up StudiesHealth Care Costs/*statistics & numerical dataHospital Costs/*statistics & numerical dataHumansMaleMiddle AgedMyocardial Infarction/*economics/*therapyMyocardial Revascularization/*economicsOutcome and Process Assessment (Health Care)Platelet Aggregation Inhibitors/economics/*therapeutic usePlatelet Glycoprotein GPIIb-IIIa Complex/antagonists & inhibitorsQuality-Adjusted Life YearsResearch Support, Non-U.S. Gov'tTyrosine/*analogs & derivatives/economics/*therapeutic useUnited Stateshttp://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=12377083[32]. At 6 months, which was the duration of the trial, the early interventional strategy was dominated by the conservative strategy. In the long-term extrapolation, 6-month survivors, with or without a myocardial infarction, was assigned life expectancy employing data from Framingham and Duke Cardiovascular Disease Database. In this analysis, the cost per life-year gained varied between $9,500 (approximately 5,200) and $17,200 (9.500) (2000 price level) using different assumptions of long-term treatment effect and different sources of long-term life expectancy. The FRISC II trial reported high incremental costs per prevented death or MI (649,000 Swedish kronor [SEK] or approximately 48,000, 2000 price level) in a within-trial analysis of one year duration ADDIN EN.CITE Janzon20024930493117413602312002JanCost-effectiveness of an invasive strategy in unstable coronary artery disease; results from the FRISC II invasive trial. The Fast Revascularisation during InStability in Coronary artery disease31-40Institution of Medicine and Care, Linkoping University, SwedenJanzon, M.Levin, L. A.Swahn, E.Eur Heart JAgedAmbulatory Care/economicsComparative StudyCoronary Arteriosclerosis/complications/*economics/mortalityCost-Benefit Analysis/economicsElectrocardiography/economicsEndpoint DeterminationFemaleFibrinolytic Agents/economics/therapeutic useFollow-Up StudiesHealth Care CostsHospitalization/economicsHumansMaleMyocardial Infarction/complications/economics/mortalityPlatelet Glycoprotein GPIIb-IIIa Complex/economics/therapeutic useProspective StudiesResearch Support, Non-U.S. Gov'tScandinavia/epidemiologySensitivity and SpecificitySurvival AnalysisVentricular Function, Left/physiologyhttp://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=11741360[33]. In a second economic evaluation from the FRISC II trial, data on costs and quality of life at two years follow-up was available and long-term costs and QALYs were extrapolated to a lifetime perspective ADDIN EN.CITE Janzon4510451Janzon, M.Avhandling[34]. The reported cost per QALY was 55,000 SEK (4,100) (2002 price level). Exploring cost-effectiveness in different sub-groups did not alter the conclusions that an early interventional strategy is cost-effective although the ICER was substantially higher for women than for men.
In general, the results from these previous studies follow a similar trend to the findings in the present analysis, namely that an early interventional strategy does not appear cost-effective in the limited short term perspective, whereas in the more relevant long-term perspective an early interventional strategy can yield positive health outcomes at an acceptable costs, at least in patients at high risk. In contrast to previous studies, the present analysis appears to demonstrate that risk stratification is likely to have a significant impact on the cost-effectiveness of an early interventional strategy. This is true when including and varying, the baseline risk and then applying a common treatment effect and when treatment effect is allowed to vary according to the baseline risk.
Sensitivity scenarios indicated that the base-case results appeared robust to the assumptions required for the long-term extrapolation with the exception of the duration of the treatment effect after the 5 year trial follow-up. Extending the duration of the long-term treatment effect of an early interventional strategy beyond the five years of follow-up observed in RITA 3 expectedly had a favourable effect on cost-effectiveness. The extent to which this may change the decision regarding the adoption of an early interventional strategy will depend upon the strength of the decision makers beliefs about the duration of the treatment effect, and whether the treatment effect is considered to be constant or is likely to vary across different risk groups.
A limitation of the present study is that comprehensive data on costs was only collected up to one year after randomization in RITA 3. However, the assumptions for the long-term extrapolation were conservative for the early interventional strategy. A further limitation is that individual-patient data was only available from RITA 3 for the present analysis. Future work should explore the impact of the results of employing individual-patient data from several studies.
In conclusion, the results of the present analysis show that an early interventional strategy in patients presenting with NSTE-ACS improves quality-adjusted survival at an acceptable cost for patients at high risk but is unlikely to do so for patients at low risk.
References
ADDIN EN.REFLIST 1. Mehta SR, Cannon CP, Fox KA, Wallentin L, Boden WE, Spacek R, Widimsky P, McCullough PA, Hunt D, Braunwald E, Yusuf S. Routine vs selective invasive strategies in patients with acute coronary syndromes: a collaborative meta-analysis of randomized trials. JAMA 2005;293:2908-17.
2. Fox KA, Poole-Wilson P, Clayton TC, Henderson RA, Shaw TR, Wheatley DJ, Knight R, Pocock SJ. 5-year outcome of an interventional strategy in non-ST-elevation acute coronary syndrome: the British Heart Foundation RITA 3 randomised trial. Lancet 2005;366:914-20.
3. Kim J, Henderson RA, Pocock SJ, Clayton T, Sculpher MJ, Fox KA. Health-related quality of life after interventional or conservative strategy in patients with unstable angina or non-ST-segment elevation myocardial infarction: one-year results of the third Randomized Intervention Trial of unstable Angina (RITA-3). J Am Coll Cardiol 2005;45:221-8.
4. Epstein D, Sculpher M, Clayton T, Henderson RA, Pocock SJ, Buxton M, Fox KA. A strategy of early angiography compared to conservative management in non-ST-elevation myocardial infarction: cost results from the third randomised intervention treatment in angina (RITA-3) trial. Submitted 2006.
5. National Institute for Clinical Excellence (NICE). Guide to the Methods of
Technology Appraisal. London: NICE, 2004.
6. Sonnenberg FA, Beck JR. Markov models in medical decision making: a practical guide. Medical Decision Making 1993;13:322-38.
7. Briggs A, Sculpher M. An introduction to Markov modelling for economic evaluation. Pharmacoeconomics 1998;13:397-409.
8. Fox KA, Poole-Wilson PA, Henderson RA, Clayton TC, Chamberlain DA, Shaw TR, Wheatley DJ, Pocock SJ. Interventional versus conservative treatment for patients with unstable angina or non-ST-elevation myocardial infarction: the British Heart Foundation RITA 3 randomised trial. Randomized Intervention Trial of unstable Angina. Lancet 2002;360:743-51.
9. Collet D. Modelling survival data in medical research. London: Chapman & Hall; 1994.
10. Government Actuary Department. Interim life tables. Expectation of life for males in the United Kingdom, based on data for the years 2002-2004. HYPERLINK "http://www.gad.gov.uk" http://www.gad.gov.uk(21 Oct 2006).
11. National Statistics. Review of the registrar general on deaths by cause, sex and age, in England and Wales. London: National Statistics, 2003.
12. Brooks R. EuroQol: the current state of play. Health Policy 1996;37:53-72.
13. Dolan P. Modeling valuations for EuroQol health states. Medical Care 1997;35:1095-108.
14. Karlsson G, Johannesson M. The Decision Rules of Cost-Effectiveness Analysis. Pharmacoeconomics 1996;9:113-120.
15. Claxton K. The irrelevance of inference: a decision-making approach to the stochastic evaluation of health care technologies. J Health Econ 1999;18:341-64.
16. Claxton K, Sculpher M, Drummond M. A rational framework for decision making by the National Institute For Clinical Excellence (NICE). Lancet 2002;360:711.
17. Fenwick E, O'Brien BJ, Briggs A. Cost-effectiveness acceptability curves--facts, fallacies and frequently asked questions. Health Econ 2004;13:405-15.
18. Spacek R, Widimsky P, Straka Z, Jiresova E, Dvorak J, Polasek R, Karel I, Jirmar R, Lisa L, Budesinsky T, Malek F, Stanka P. Value of first day angiography/angioplasty in evolving Non-ST segment elevation myocardial infarction: an open multicenter randomized trial. The VINO Study. Eur Heart J 2002;23:230-8.
19. Anderson HV, Cannon CP, Stone PH, Williams DO, McCabe CH, Knatterud GL, Thompson B, Willerson JT, Braunwald E. One-year results of the Thrombolysis in Myocardial Infarction (TIMI) IIIB clinical trial. A randomized comparison of tissue-type plasminogen activator versus placebo and early invasive versus early conservative strategies in unstable angina and non-Q wave myocardial infarction. J Am Coll Cardiol 1995;26:1643-50.
20. Lagerqvist B, Husted S, Kontny F, Stahle E, Swahn E, Wallentin L. 5-year outcomes in the FRISC-II randomised trial of an invasive versus a non-invasive strategy in non-ST-elevation acute coronary syndrome: a follow-up study. Lancet 2006;368:998-1004.
21. de Winter RJ, Windhausen F, Cornel JH, Dunselman PH, Janus CL, Bendermacher PE, Michels HR, Sanders GT, Tijssen JG, Verheugt FW. Early invasive versus selectively invasive management for acute coronary syndromes. N Engl J Med 2005;353:1095-104.
22. Boden WE, O'Rourke RA, Crawford MH, Blaustein AS, Deedwania PC, Zoble RG, Wexler LF, Kleiger RE, Pepine CJ, Ferry DR, Chow BK, Lavori PW. Outcomes in patients with acute non-Q-wave myocardial infarction randomly assigned to an invasive as compared with a conservative management strategy. Veterans Affairs Non-Q-Wave Infarction Strategies in Hospital (VANQWISH) Trial Investigators. N Engl J Med 1998;338:1785-92.
23. McCullough PA, O'Neill WW, Graham M, Stomel RJ, Rogers F, David S, Farhat A, Kazlauskaite R, Al-Zagoum M, Grines CL. A prospective randomized trial of triage angiography in acute coronary syndromes ineligible for thrombolytic therapy. Results of the medicine versus angiography in thrombolytic exclusion (MATE) trial. J Am Coll Cardiol 1998;32:596-605.
24. Cannon CP, Weintraub WS, Demopoulos LA, Vicari R, Frey MJ, Lakkis N, Neumann FJ, Robertson DH, DeLucca PT, DiBattiste PM, Gibson CM, Braunwald E. Comparison of early invasive and conservative strategies in patients with unstable coronary syndromes treated with the glycoprotein IIb/IIIa inhibitor tirofiban. N Engl J Med 2001;344:1879-87.
25. Sutton AJ, Abrams KR, Jones DR, Sheldon T, Song F. Methods for Meta-Analysis in Medical Reseach. Chichester: Wiley; 2000.
26. Briggs A, Sculpher M, Claxton K. Decision modelling for health economic evaluation: Oxford University Press, USA; 2006.
27. Greene W. Econometric Analysis. 5th ed. New York: Prentice Hall; 2003.
28. van Domburg RT, van Miltenburg-van Zijl AJ, Veerhoek RJ, Simoons ML. Unstable angina: good long-term outcome after a complicated early course. J Am Coll Cardiol 1998;31:1534-9.
29. STATA version 7. Stata Statistical Software: Release 7.0. College Station, TX: Stata Corporation.
30. Heart Protection Study Collaborative Group. Lifetime cost effectiveness of
simvastatin in a range of risk groups and age groups derived from randomised trial of
20536 people. BMJ 2006;333:1145-8.
31. Briggs A, Mihaylova B, Sculpher M, Hall A, Wolstenholme J, Simoons M, Deckers J, Roberto F, Remme WJ, Bertrand M, Fox K. The Cost-effectiveness of perindopril in reducing cardiovascular events in patients with stable coronary artery disease using data from the EUROPA Study. Heart published online 29 Nov 2006;
doi:10.1136/hrt.2005.086728
32. Mahoney EM, Jurkovitz CT, Chu H, Becker ER, Culler S, Kosinski AS, Robertson DH, Alexander C, Nag S, Cook JR, Demopoulos LA, DiBattiste PM, Cannon CP, Weintraub WS. Cost and cost-effectiveness of an early invasive vs conservative strategy for the treatment of unstable angina and non-ST-segment elevation myocardial infarction. JAMA 2002;288:1851-8.
33. Janzon M, Levin LA, Swahn E. Cost-effectiveness of an invasive strategy in unstable coronary artery disease; results from the FRISC II invasive trial. The Fast Revascularisation during InStability in Coronary artery disease. Eur Heart J 2002;23:31-40.
34. Janzon M. Treatment strategies in unstable coronary artery disease, economic and quality of life evaluations [dissertation]. Medical dissertation No. 811: Linkping university; 2003.
35. Claxton K, Sculpher M, McCabe C, Briggs A, Akehurst R, Buxton M, Brazier J, O'Hagan T. Probabilistic sensitivity analysis for NICE technology assessment: not an optional extra. Health Econ 2005;14:339-47.
36. Stinnett AA, Paltiel AD. Estimating CE Ratios under Second-order Uncertainty: The Mean Ratio versus the Ratio of Means. Medical Decision Making 1997;17:483-489.
Appendix
Details of the probabilistic assessment
The model was evaluated for 60 Markov cycles (years) implying that all hypothetical individuals would, in effect, be in the Dead state at the termination of the analysis regardless of the chosen starting age of the cohort. In each simulation in the probabilistic analysis ADDIN EN.CITE Claxton20053760376157361421442005AprProbabilistic sensitivity analysis for NICE technology assessment: not an optional extra339-47Centre for Health Economics, Department of Economics and Related Studies, University of York, UK. kpcl@york.ac.ukClaxton, K.Sculpher, M.McCabe, C.Briggs, A.Akehurst, R.Buxton, M.Brazier, J.O'Hagan, T.Health EconCost-Benefit AnalysisDecision Support Techniques*Guidelines*Models, StatisticalTechnology Assessment, Biomedical/economics/*standards/*statistics &numerical datahttp://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=15736142[35], parameter values were drawn randomly from the defined probability distributions and the cohort of hypothetical individuals was run through the model and mean costs and health outcomes calculated for both strategies. This procedure was repeated 5 000 times generating 5 000 estimates of mean costs and mean effects for both strategies. The expected costs and effects for each treatment strategy are the mean of these 5000 simulations ADDIN EN.CITE Stinnett19972680268Stinnett, A. A.Paltiel, A. D.1997Estimating CE Ratios under Second-order Uncertainty: The Mean Ratio versus the Ratio of MeansMedical Decision Making17483-489[36]. The expected costs and effects are then combined to an incremental cost-effectiveness ratio.
As the statistical models were based on individual patient data it was possible to estimate correlations between parameters as well as means and standard errors to be employed in the probabilistic analysis. The Cholesky decomposition matrix, T, is derived from the variance- covariance matrix, V, such that TT=V. A vector of correlated parameters, x, with variance and covariance corresponding to the variance-covariance matrix, can be estimated from the following equation x=y+Tz, where y is the vector of parameter means and z is a vector of independent random draws from the standard normal distribution ADDIN EN.CITE Briggs20064101410Briggs, A.Sculpher, M.Claxton, K.2006Decision modelling for health economic evaluationOxford University Press, USAGreene20033741374Greene, WH.2003Econometric AnalysisNew YorkPrentice Hall5th[26,27]. The distributional assumptions employed in the analysis were multivariate normality of the log-odds scale for the logistic models, multivariate normality of the log hazard scale for the survival-analysis model, and multivariate normality on the raw cost and HRQOL scales for costs and QALYs, respectively. Cholesky decomposition matrixes for the equations reported in the data section are shown below.
Cholesky decomposition matrixes
Table A1. Equation 1
TreatAgeAnginaConstantTreat0.2882000Age0.00090.161200Angina0.0007-0.01140.28330Constant-0.1724-0.2023-0.14520.1416
Table A2. Interaction model index hospitalisation
TreatRiskTreat*riskConstantTreat0.5390.00Risk0.7910.96100Treat*risk-1.468-0.4030.8410Constant-0.309-0.189-0.1240.142
Table A3. Equation 2
AgeDiabetesPrevious MISmokerPulseST-
depressionAnginaMaleLeft BBBTreatConstant l n A g e 0 . 0 8 7 0 0 0 0 0 0 0 0 0 0 0 D i a b e t e s - 0 . 0 0 4 0 . 1 7 3 0 0 0 0 0 0 0 0 0 0 P r e v i o u s M I - 0 . 0 1 2 - 0 . 0 1 4 0 . 1 5 3 0 0 0 0 0 0 0 0 0 S m o k e r 0 . 0 5 1 0 . 0 0 8 0 . 0 0 7 0 . 1 5 1 0 0 0 0 0 0 0 0 P u l s e 0 . 0 0 1 - 0 . 0 0 5 - 0 . 0 0 0 0 . 0 0 0 0 . 0 2 4 0 0 0 0 0 0 0 S T d e p r e s s i o n - 0 . 0 2 1 - 0.0060.019-0.002-0.0210.144000000Angina0.003-0.002-0.023-0.0030.006-0.0030.14700000Male0.004-0.005-0.021-0.0090.0170.0000.0210.1540000Left BBB-0.033-0.004-0.0280.010-0.0390.006-0.0080.0010.261000Treat-0.000-0.005-0.0050.0080.0030.001-0.001-0.0010.0060.14800Constant-0.1220.007-0.041-0.049-0.197-0.069-0.077-0.104-0.026-0.0560.0920 l n 0 . 0 0 3 0 . 0 0 2 0 . 0 0 1 0 . 0 0 0 0 . 0 0 0 0 . 0 0 2 0 . 0 0 1 0 . 0 0 1 0 . 0 0 2 - 0 . 0 0 2 - 0 . 0 4 3 0 . 0 5 4
T a b l e A 4 . I n t e r a c t i o n m o d e l f o l l o w - u p p e r i o d
T r e a t R i s k T r e a t * r i s k C o n s t a n t L n g a m m a T r e a t 0 . 2 7 8 6 0 0 0 0 R i s k 0 . 2 4 5 7 0 . 4 0 7 0 0 0 0 T r e a t * r i s k - 0 . 7 4 1 5 - 0 . 1 0 7 3 0 . 4 5 7 1 0 0 C ! " # $ uuuuuk^P h hO H*OJ QJ ^J hO hO OJ QJ ^J hO OJ QJ ^J hO hO 5H*OJ QJ ^J h hO 5H*OJ QJ ^J h hO 5OJ QJ ^J hj hO 5H*OJ QJ ^J hj hO 5OJ QJ ^J hj hO OJ QJ ^J #hj hO 5CJ OJ QJ ^J aJ h+ hO OJ QJ ^J #hO hO 5CJ OJ QJ ^J aJ ! " # X
b
gdO $a$gdO y z )z $ X Y
!
b
c
# γzobXboJ h+ h_ 5OJ QJ ^J h_ OJ QJ ^J h+ h_ OJ QJ ^J h_ 5OJ QJ ^J #h+ hr 5CJ OJ QJ ^J aJ ho 5CJ OJ QJ ^J aJ hr hH hr hs} hO OJ QJ ^J hP2 hO OJ QJ ^J h+ hO 5OJ QJ ^J h+ hO OJ QJ ^J h OJ QJ ^J h hO H*OJ QJ ^J h hO OJ QJ ^J
# R w F
. ` s ^ _ ` dh gdr d gdO dh gd_ dh gdO gdr gdO
F
M
. 6 7 S ; ` h l q 2 I [ \ r s ~ ^ _ ` i j 鲥z #h hO 5CJ OJ QJ ^J aJ h+ hO 5OJ QJ ^J h_ 5OJ QJ ^J h_ h<9 OJ QJ ^J h;
h_ OJ QJ ^J h+ h_ 5OJ QJ ^J h2 h_ 5OJ QJ ^J h h_ 5OJ QJ ^J h_ OJ QJ ^J h+ h_ OJ QJ ^J *j 7 8 r s v w z { ! ! ! ! " " ( ( ( ( ( ( c) d) U2 V2 [2 \2 Q3 ˷}qe h;zN CJ OJ QJ ^J h_ CJ OJ QJ ^J hf CJ OJ QJ ^J %j hr hr CJ OJ QJ U^J hxc CJ OJ QJ ^J hr hr CJ OJ QJ ^J &hr hr 5CJ OJ QJ S*^J aJ +hr hr 5CJ OJ QJ ^J aJ mH sH hWE h+ hO 5OJ QJ ^J h+ hO OJ QJ ^J % ( ( Q3 R3 r; s; t; ~; ; > > \A ]A wA yA A D D G G =S >S dh gdv` d gd04u dh gd04u dh gd dh gdf dh gd;zN dh gd;zN dh gdr Q3 R3 4 5
5 5 ; ; ;
; r; s; t; u; |; }; ~; ; ; ; <