Overview

A probabilistic decision analytic model was developed to assess the cost-effectiveness of EECP in the UK NHS. A systematic review was conducted to obtain effectiveness evidence from randomized control trials (RCTs) (Reference McKenna, McDaid and Suekarran17). The model provides a framework for the synthesis of evidence identified from the review and the elicitation of unknown parameters from experts. The model considers the potential long-term costs and benefits associated with the outcome of improvement in health-related quality of life (HRQoL) from EECP.

The model evaluates costs from the perspective of the National Health Service and Personal Social Services (NHS & PSS), expressed in UK £ sterling at a 2008 price base. Outcomes in the model were expressed in terms of quality-adjusted life-years (QALYs), with costs and benefits discounted at 3.5 percent per year (19).

Treatment Strategies and Population

The decision model evaluates the cost-effectiveness of EECP in adults with chronic stable angina. It evaluates a strategy of EECP treatment compared with no treatment on the assumption that angina patients would receive EECP treatment over and above standard current clinical practice care. This is consistent with the one RCT (MUST-EECP trial) assessing EECP in angina, which compares EECP treatment with sham-EECP (Reference Arora, Chou and Jain3). The base–case population in the model relates to the baseline characteristics of this trial population, under the assumption that this population is representative of angina patients typically presenting for EECP in the current clinical setting.

Model Structure

The decision model is structured to capture the HRQoL benefits and costs associated with the treatment strategies up to a period of 12 months, and to project the health related benefits beyond this 1-year period to a lifetime time horizon. Given that there is no evidence to suggest that EECP treatment compared with placebo has a differential impact on the risk of experiencing nonfatal-cardiovascular disease (CVD) events and death, it is assumed that the benefits of EECP are purely palliative. The model incorporates the risk of CVD events and death to keep track of the number of patients alive who could potentially benefit from QOL improvements, but because there is no differential impact on total costs and benefits between the treatment strategies due to CVD events, the estimate of cost-effectiveness is not affected by these events.

In the first year after EECP treatment, patients are assumed to achieve, on average, the HRQoL benefits reported in the 12-month follow-up of the MUST-EECP trial (Reference Arora, Chou and Jain4). A Markov state transition model (Reference Briggs, Claxton and Sculpher6) (Supplementary Figure 1, which is available at www.journals.cambridge.org/thc2010015) was developed based on the likelihood of sustaining HRQoL benefits over time. Three health states were defined as “Responders,” “Nonresponders,” and “Dead.” The “Responders” state represents patients who sustain HRQoL benefits from EECP. The “Nonresponders” state represents patients who lose initial HRQoL benefits from EECP. The “Dead” state incorporates death from cardiovascular causes and from other causes. Given that patients, on average, achieve the HRQoL benefits reported in the MUST-EECP trial at the end of the first year, patients in the second year either continue to sustain the 1-year HRQoL benefits and will move to the “Responders” state, or they will lose the benefits by falling back to baseline HRQoL and entering the “Nonresponders” state. Consideration was given to adding an additional state representing partial responders, that is, those patients who partially sustain HRQoL benefits, where their HRQoL is better than baseline but worse than 12 months after EECP. However, due to a lack of any evidence to populate this transition, or to determine the HRQoL of these patients, it was excluded from the model. In the third and subsequent years, responders to EECP in the previous year will either continue to sustain their HRQoL benefits or enter the “Nonresponders” state. Recovery of response in each year was possible if patients received repeat top-up procedures. All patients in the model face a risk of cardiovascular events, which eventually lead to death, and are deemed to be at a competing risk of a noncardiovascular death.

Model Inputs

Clinical Effectiveness. The systematic review identified only one RCT (the MUST-EECP trial) comparing EECP with an alternative treatment strategy (Reference Arora, Chou and Jain3;Reference Arora, Chou and Jain4). Full details of the systematic review undertaken and quality assessment of the identified studies are reported elsewhere (Reference McKenna, McDaid and Suekarran17). The one identified RCT on EECP was considered to be of good quality. A total of 139 participants were randomized: 72 to active-EECP and 67 to inactive-EECP. Of the participants in the active-EECP arm, thirteen withdrew during follow-up mainly due to adverse events, while only two withdrew from the inactive-EECP arm (Reference Arora, Chou and Jain3). All aspects of clinical effectiveness reported in this trial were considered for inclusion in the cost-effectiveness model including (i) exercise treadmill duration; (ii) time to ≥1 mm ST-segment depression; (iii) angina counts; (iv) nitroglycerin use; and (v) HRQoL. Despite additional searches, no evidence was available to link the four intermediate outcomes to final health outcomes in terms of QALYs. As a result, the primary outcome used in the model was improvement in HRQoL as reported in the trial itself. At 12 months after the end of treatment, the trial reports improvement in HRQoL from baseline assessed by the SF-36 instrument (Reference Arora, Chou and Jain4).

Health-Related Quality of Life. To estimate QALYs, it is necessary to quality-adjust the period of time the average patient is alive within the model using an appropriate HRQoL weight (utility). The trial reported summary measures of improvements in HRQoL from baseline to 12 months in the eight dimensions of the SF-36 in both the EECP and sham-EECP treatment arms (Reference Arora, Chou and Jain4). A recently developed algorithm was applied to predict a preference-based EQ-5D utility score using the summary scores of the eight SF-36 dimensions because patient level data were not available (Reference Ara and Brazier2). These utility values were used to estimate the incremental change in utility for EECP relative to no treatment over a 12-month period. To incorporate uncertainty in the estimate of the incremental change in utility in the absence of details of sampling uncertainty in the trial, a beta distribution was applied with a standard error equivalent to half the mean change in utility (Reference Briggs, Goeree, Blackhouse and O'Brien8). Table 1 reports the utility improvement for EECP relative to no treatment at 1-year. Baseline utility values were taken from age- and sex-dependent population norms for the general UK population and adjusted downward to reflect the presence of angina in this population (Reference Kind, Hardman and Macran14).

Table 1.

Utility Improvement for EECP Relative to No Treatment at 1 Year

EECP, enhanced external counterpulsation; SE, standard error assumed equivalent to half the mean change in utility.

Duration of HRQoL Benefits. Beyond 12 months, there is no trial evidence examining the degree to which improvement in HRQoL from EECP is sustained over time. In the absence of suitable trial estimates, a wider set of studies incorporating the experience of patients in the International EECP Patient Registry (IEPR) were examined. However, despite examining all published studies in EECP, the review did not identify a single source that could provide a generic measure of HRQoL beyond 12 months. To quantify the treatment duration in terms of sustaining HRQoL benefits, formal techniques were used to elicit the opinions of clincal experts. This involved asking clinical experts to report their beliefs about the duration of HRQoL benefits with some estimate of their uncertainty (Reference O'Hagan, Buck and Daneshkhah20). Experts were told to assume that patients have any additional repeat procedures (or top-up sessions) when they require them. An Excel-based exercise was designed to elicit the probability of sustaining benefits in subsequent years. One repetitive question was asked throughout the exercise; for example, in the second year: “In year 2, what proportion of patients would you expect to sustain the average HRQoL benefits observed at one year following end of treatment?”

Five experts with experience and knowledge of EECP in the United Kingdom completed the exercise independently giving their own belief about the unknown quantities with estimates of uncertainty. The format chosen for each of the questions was a frequency chart (Reference van Noortwijk, Dekker, Cooke and Mazzuchi25). Experts were asked to place 20 crosses on the frequency chart to represent their current belief and uncertainty about that particular question. A distribution of uncertainty for the parameters was then derived. Full details of the elicitation exercise are reported elsewhere (Reference McKenna, McDaid and Suekarran17). Mean and standard deviations for the probability of sustaining HRQoL benefits in each subsequent year are shown in Table 2. The results from each expert were linearly pooled and a beta distribution applied to the values (Reference Cooke9). Each expert was given equal weight and the pooled result was assumed to be representative of the beliefs of relevant clinical experts.

Table 2.

Mean and SD of the Elicited Values for Each Expert Separately and Linearly Pooled Results Across Experts

^aConditional on sustaining benefits in the previous year, and conditional on receiving top-up procedures as considered appropriate.

SD, standard deviation; QoL, quality of life.

Mortality. The model separates deaths into those caused by CVD events and other cause mortality, although no treatment effect from EECP in terms of mortality was modeled. The mortality associated with CVD events was informed by the risk equations applied in the EUROPA trial (Reference Briggs, Mihaylova and Sculpher7). The output was the number of deaths from CVD causes in each yearly cycle based on the likelihood of a first primary event of CVD death, nonfatal myocardial infarction (MI) or cardiac arrest, and the likelihood of subsequent CVD deaths from nonfatal primary events.

The age-dependent risk of other-cause mortality was based on standard UK age- and sex-specific mortality rates (11). These were adjusted to exclude those deaths recorded with an International Classification of Diseases code pertaining to cardiovascular disease (ICD-10 I20-I99). The treatments were assumed not to infer a differential mortality effect.

Resource Use and Unit Costs. Resource utilization and cost data were based on treatment received. Because no additional costs are incurred under a no-treatment strategy the only costs included in the model were those associated with EECP. The costs associated with EECP relate to the standard 35-hour treatment sessions and the need for repeat top-up procedures over time.

The cost of EECP per patient was based on the average number of patients per annum that a UK center can currently take and the cost of consumables. UK centers currently run at approximately 12 patients per annum (Ken Miles and Wayne Sheedy, personal communications, 2008) and is limited by referral rates. The capital cost of a new EECP machine (“AngioNew”) was taken to be £90,000 +VAT (including installation and training for three therapists for 3 days) (Ken Miles, personal communication, 2008). The machine is expected to have a useful life of approximately 10 years, which gives an equivalent annual cost of £10,822 (with an annuity factor for 10 years included at an interest rate of 3.5 percent per annum). Typical equipment replacement costs include 1 or 2 sets of cuffs per year, 1 set of hoses per year, and replacement of the plethysmograph every 2 years. The unit costs associated with each of these were informed by Vasogenics current price list (effective from April 2007). The consumables per patient for all thirty-five sessions are typically one pair of trousers, pre-assessment ultrasound scan, gel, and electrocardiography electrodes. Table 3 provides a breakdown of the total per patient cost for 35-hours of EECP treatment. Allowing for overheads and staffing costs, the total per patient cost was estimated to be £4,347 per treatment over a 35-hour course. Some patients may receive a repeat or top-up procedure involving fewer treatment sessions. The cost per session was obtained by dividing the total cost of treatment by thirty-five, giving a cost of £124 per session. The additional cost of repeat procedures was based on an average of 10 additional sessions.

Table 3.

Resource Use and Unit Cost Inputs Used in the Model

ECG, electrocardiography; FTE, full-time employment.

Repeat Procedures. A typical course of EECP involves a total of 35 hours of therapy. Some patients require (or request) a repeat or top-up procedure involving several additional sessions. These sessions are generally given to help sustain the long-term benefits of EECP. A search of the literature was undertaken to identify studies that could potentially inform the rate of repeat or top-up procedures. The search identified one study, based on the experience of patients in the IEPR, which examined the frequency and efficacy of repeat EECP for stable angina (Reference Michaels, Barsness and Soran18). Within 2 years of the initial course of EECP, the rate of repeat EECP was 18 percent (194/1078 patients), which occurred at a mean interval of 378 days after initial EECP. Assuming a fixed rate with respect to time, the 2-year probability was converted to a 2-year rate and used to generate an annual probability (Reference Briggs, Claxton and Sculpher6). This annual probability of repeats decreased exponentially over time. To incorporate uncertainty in the estimates of repeat procedures, a beta distribution was applied.

Analytical Methods. Probabilistic analyses were conducted using Monte Carlo simulation (repeated random sampling from the joint probability distribution of parameters). The results are presented in two ways. First, mean lifetime costs and QALYs of both treatment strategies are presented and their cost-effectiveness compared using an incremental cost-effectiveness ratios (ICER) (Reference Johannesson and Weinstein12). Second, decision uncertainty is presented as the probability that each strategy is considered the more cost-effective option for a given cost-effectiveness threshold. Two alternative thresholds are used, £20,000 and £30,000 per QALY gained, reflecting the range used by the National Institute for Health and Clinical Excellence (NICE) (Reference Devlin and Parkin10;19).

The following analyses are undertaken for an average starting age of 64 years and 92 percent of subjects male (based on the patient characteristics of the MUST-EECP trial) (Reference Arora, Chou and Jain4). First, separate analyses are undertaken by assuming that HRQoL benefits from EECP are sustained for different durations: (i) a worst case scenario is considered where HRQoL gains from EECP are only maintained in the first year after treatment, and lost in subsequent years; (ii) a best case scenario is considered where QoL benefits are assumed to last over a patient's lifetime; and (iii) a more realistic scenario is considered where the cost-effectiveness is examined in terms of the proportion of patients likely to sustain benefits over time. The third scenario based on the pooled consensus of experts forms the base–case analysis. In an alternative scenario, the empirical values from each expert are considered separately.

Second, the impact of the cost of EECP on cost-effectiveness is examined. The base–case assumes a total cost of £4,347 per patient as explained above. In an alternative scenario, the cost of EECP is increased/decreased to reflect the possibility of increased/decreased utilization (patient throughput) in some centers. Referral restrictions currently limit the throughput of patients undergoing EECP in NHS centers. Third, the rate of repeat EECP sessions is examined. The base–case assumes that the probability of repeat or top-up EECP sessions is 18 percent within 2 years of initial treatment. In an alternative scenario, this probability is varied from 10 percent to 30 percent.