Results of Economic Literature Review
Figure 1 presents a flow chart describing the process of identifying relevant literature. A total of three economic evaluations were included in the review.[3,4,105] A summary of each study is presented in Table 1.
Studies eliminated/selected for the review after applying the inclusion/exclusion criteria.
General Characteristics of the Studies Included
All studies were published between 2007 and 2008. Two of the studies were set in the UK[4,105] and one was in the USA. All three studies used Markov models to represent the clinical pathway using a lifetime horizon. Two used a NHS perspective,[4,105] while the third used a societal perspective (although only direct healthcare costs were included). One study was used to inform guidelines on lipid-lowering therapies in the UK and is referred to as 'Guideline' in the current document. Two studies discounted future costs and benefits at 3.5%,[4,105] while the third used 3% discount rates.
All three studies compared statin monotherapies. Chan and Guideline[3,4,105] obtained the relative risks (RR) of events (Table 2) by combining data from Aggrastat to Zocor (AtoZ; simvastatin 80 mg/day vs simvastatin 20 mg/day) and Pravastatin or Atorvastatin Evaluation and Infection Therapy (PROVE-IT; atorvastatin 80 mg/day vs pravastatin 40 mg/day).[5,6] Chan assumed that individuals with ACS did not remain at high risk for life and after the first 2 years applied RRs estimated from individuals with stable coronary artery disease (CAD) enrolled in Incremental Decrease in End Points Through Aggressive Lipid Lowering (IDEAL; atorvastatin 80 mg/day vs simvastatin 20(40) mg/day) and Treating to New Targets (TNT) (atorvastatin 80 mg/day vs atorvastatin 10 mg/day).[7,8] Ara used the results of a Bayesian mixed treatment comparison of data from 28 studies involving atorvastatin 80 mg/day, rosuvastatin 40 mg/day, simvastatin 80 mg/day, simvastatin 40 mg/day and placebo. The RRs of cardiovascular events (comparing the individual statins with placebo), were estimated using the observed changes in LDL-C and a published relationship between LDL-C and cardiovascular events.
Health States Modeled
When defining health states for economic models, there is a trade-off between replicating the exact clinical pathway and using the most clinically important health states. The former increases the complexity and computational running time of the model and may not necessarily add anything to the latter. In reality, the number of health states modeled is frequently governed by the evidence available to populate them. It is immediately apparent that there is a large difference in the number and type of health states used in the studies included in this review (Table 2). Chan supported their choice of health states as being components of the composite end point in at least one of the four clinical trials used for effectiveness, while guidelines included all types of cardiovascular events reported in the studies. Ara was governed by the evidence available to translate changes in LDL-C to the RRs of events.
Fatal Health States Chan did not differentiate between CVD and non-CVD fatalities and assumed that all-cause mortality (ACM) was 1.16-times higher than the US general population rates (Table 2). Guideline had two separate health states for fatalities (CVD and non-CVD) and assumed that the risk of non-CVD death was twice as high as the UK general population rates.[10,105] Ara had three separate health states for fatalities (CHD, CVD [vascular but noncardio] and non-CVD). Chan applied the ACS ACM RR (0.76) during the first 2 years followed by the CAD ACM RR (0.99) to all fatalities. Guideline applied the ACM RR to fatal CVD events (0.76) and assumed a RR of 1 for non-CVD fatalities. Ara also assumed a RR of 1 for non-CVD fatalities and applied different RRs for CHD fatalities and CVD (vascular but noncardio) fatalities (Table 2). As a substantial proportion of benefits in CVD models are accrued from deaths avoided, the different approaches could potentially have a large effect on the results.
Nonfatal Health States Individuals in Chan's model commenced in the 'remains well' health state and, during each annual cycle, could experience a myocardial infarction (MI), stroke, rehospitalization for angina, revascularization or death. Individuals in Guideline's model commenced in the 'well post a primary event' health state and, during each annual cycle, could experience a MI, revascularization, unstable angina, heart failure, stroke or death. Individuals in Ara's model commenced in one of three qualifying event health states: revascularization, unstable angina or MI (all during the previous 28 days). During each annual cycle, individuals could experience rehospitalization for unstable angina, a MI, stroke or death. For individuals not experiencing a subsequent event in the current cycle, Ara and Guideline both modeled postevent health states for all nonfatal events.[4,105] Ara also included several duplicate health states defined to retain the memory of previous events, such as rehospitalization for angina given a history of a stroke.
Transitions between Health States
Both Chan and Guideline used the event rates observed in the lower dose statin arms in the clinical studies.[3,105] Chan applied constant annualized event rates from all health states, with ACS data used in the first 2 years followed by data from the two CAD studies. Guideline applied the observed event rates from the two ACS clinical studies to individuals in the 'well post primary event' health state, assuming 40% of events occurred in the first year. Transitions to subsequent events were obtained from the literature and, with the exception of transitions from heart failure and the revascularization health states, the rates were adjusted for age. As some of these data were obtained from individuals not receiving statin treatment, the number of events in the low-dose statin arm and, consequently, the benefits of treatment, may have been overestimated. Ara's model is slightly different in that the baseline is no treatment as opposed to a low-dose statin regimen. All transitions for the no treatment arm were obtained from individuals in the UK who did not receive statin treatment.[13–16]
Health-related Quality of Life
There are some differences in the health-related quality of life (HRQoL) data and the methods used to populate the models with utility data (Table 3). Chan assumed a constant baseline utility of 0.974 estimated from patients' responses to ratings on scales of 0–100.[3,17] Guideline and Ara used EQ-5D preference-based age-adjusted baseline utilities obtained from members of the general population.[4,18,105] While the baseline age-specific data were adjusted downward for the qualifying health states in one of the models, it is unclear if this was done in the other. All analysts applied long-term disutilities for individuals experiencing either a MI or a stroke but only two adjusted these upwards after the first year to characterize an improvement in QoL associated with the increased time since the acute event.[4,105] Chan assumed a short-term disutility of 1, 1 and 4 weeks for angina, angioplasty and coronary artery bypass graft, respectively, after which they reverted to the baseline utility level.
Using a constant baseline utility unadjusted for age overestimates the benefits of treatment to such an extent that it can potentially influence a policy decision based on a cost per QALY threshold. While Chan adjusts the baseline slightly to 0.974, the mean EQ-5D measure from the UK population at the age of 60 years (the starting age in Chan's model) is 0.820. Consequently, an event avoided in this model will accrue substantially more benefits in terms of QALYs than a corresponding event in the other models. Although the technique used to combine HRQoL data can have a large effect on results generated, only one study reports the method used to combine the utility data.
The transitions described in both Chan's and Guideline's studies imply that individuals could move to a health state with a higher utility score.[3,105] For example, someone with a history of stroke, which has a long-term utility of between 0.39 and 0.76 in Chan's model and 0.63 (before adjustment for age) in Guideline's model, could then experience a MI which has a long-term utility of 0.90 in Chan's model and 0.88 (before adjustment for age) in Guideline's model. Consequently, someone with a history of a MI following a stroke would have a higher HRQoL than someone with a history of just stroke, which is counter-intuitive and unlikely to be clinically correct.
All models used health-state costs obtained from the literature. In each model, a first-year event cost was followed by an annual long-term cost associated with the event.[3,4,105] Additional monitoring costs for individuals receiving the higher dose statins were applied by Ara and Guideline.[4,105] Ara and Chan both assigned health-state costs to individuals in the qualifying health state(s)[3,4] but there was insufficient detail in Guideline's report to ascertain if individuals in the qualifying event received an ongoing cost. If ongoing costs were not applied, the cost offset associated with avoiding an initial event could be overestimated. Another large difference was in the costs assigned to the fatal health state. Ara applied a cost of £592 for fatal cardiac events and £3688 for other fatal vascular events while Chan applied a cost of approximately £6449 to all fatalities irrespective of the cause of death.[3,4] Conversely, Guideline did not assign a cost to any fatalities. While it is unlikely that every individual in the model would incur a cost associated with death, it is also probable that at least a proportion of the CVD fatalities would incur healthcare costs.
The transitions described in both Chan's and Guideline's studies imply that individuals could move to a health state with a lower long-term ongoing cost.[3,105] For example, someone with a history of stroke, which has a long-term annual cost of between £1872 and £6934 in Chan's model and an annual cost of £2163 in Guideline's model, could then experience unstable angina, which has long-term annual costs of £1179 and £500 in Chan's and Guideline's models, respectively.[3,105]
The treatment costs varied substantially depending on the costs in each of the settings (Table 4). As the costs of statins were in flux when Chan conducted their evaluation, they explored the effect of using different incremental treatment costs in sensitivity analyses. To address potential changes in the cost of atorvastatin when the patent expires in 2011, Ara explored the effect on results assuming the cost of atorvastatin reduces from £368 per annum to £90, or if it decreased in line with that observed for simvastatin (i.e., to £21) per annum.
Adverse Events & Adherence to Potent-dose Statin Therapy
Although the safety of statins as a class is well reported, the safety of intensive-dose statin therapy is less clear owing to the smaller number of clinical trials using these treatments. To our knowledge there is no published evidence on adverse event rates associated with intensive-dose statin use in clinical practice. Silva et al. meta-analyzed the data (n = 27,548) from the clinical trials used to model the effectiveness of treatments in two of the economic evaluations in this review. The intensive-dose therapy (atorvastatin or simvastatin 80 mg/day) was associated with a significant increase in the risk of any adverse event (odds ratio [OR]: 1.44; 95% CI: 1.33–1.55; p < 0.001) and an increased risk for abnormalities on liver function testing (OR: 4.48; 95% CI: 3.27–6.16; p < 0.001) and elevations in creatine kinase (OR: 9.97; 95% CI: 1.28–77.92; p = 0.028). More recently, data from SEARCH showed an 11-fold increase in myopathy/myositis and defined premyositis in individuals receiving simvastatin 80 mg/day compared with simvastatin 20 mg/day.
The most common adverse event, myopathy, is likely to affect adherence rates. Much of the published evidence examines adherence to statins as a class, and data subgrouped by statin type or dose are not provided. Data from observational cohort studies suggest that the level of adherence, deemed as those taking at least 80% of therapy, outside the clinical trial setting decreases with time. Adherence rates can fall below 50% after 2 years[22–24] with the largest decline in the first 12 months, with the consequences of a profound 40% increase in cardiovascular events occuring and hospitalizations. However, these data are primarily for moderate- or low-dose statins and there are no data currently available for the higher doses (atorvastatin 80 mg/day, rosuvastatin 40 mg/day and simvastatin 80 mg/day) in routine practice. There is evidence suggesting that regular cholesterol monitoring, including follow-up lipid tests and physician visits, can improve adherence rates,[23,27,28] while being well-informed by physicians prior to initiation of secondary prevention treatment can improve continuous use of statin therapy.
None of the studies included in this review modeled adverse events explicitly and only one of the studies explored the potential effect on results when assuming different adherence levels for the more potent statins. The authors stipulated that different adherence rates would reflect the potential greater rates of discontinuation to the higher dose therapies owing to the less serious adverse events, such as myopathy or myalgia.
Expert Rev Pharmacoeconomics Outcomes Res. 2009;9(5):423-433. © 2009
Cite this: Are Intensive Lipid-lowering Regimens an Optimal Economic Strategy in Patients with ACS? An Acute and Chronic Perspective - Medscape - Oct 01, 2009.