Cost-effectiveness of transcatheter aortic valve implantation (TAVI) for aortic stenosis in patients who are high risk or contraindicated for surgery: a model-based economic evaluation

Aetiology

Aortic stenosis (AS) is a condition in which the aortic valve becomes progressively narrowed, leading to gradual obstruction of left ventricular outflow. The most common form, occurring in about 80% of cases, is degenerative. 1 This primarily presents as calcific AS2 and shares a similar developmental process and risk factors to atherosclerosis. 2–4 Risk factors for the development of AS include male sex, hypertension, elevated low-density lipoproteins, diabetes, cholesterol, smoking and a family history of heart disease. 1,5 Patients with chronic AS may remain asymptomatic for many years but, in most patients, symptoms of disease eventually develop. AS is usually considered in terms of its haemodynamic and symptomatic disease severity.

Haemodynamic diagnosis and severity measurement can be conducted using echocardiography, Doppler studies and cardiac catheterisation. 2,5 Severity can be assessed in terms of the aortic valve area (AVA), peak aortic jet velocity (V_max) and transaortic valve gradient, severe AS being defined as AVA < 1 cm², V_max > 4 m/second and mean transaortic valve gradient > 40 mmHg. 6

Clinical symptoms of AS include chest pain or breathlessness on exertion, angina, dizziness and syncope, and patients may suffer sudden death. 2,5 The New York Heart Association (NYHA) heart failure classification is used to classify the severity of symptoms, ranging from class I, in which the patient has no limitation in daily physical activity, to class IV, in which the patient is breathless at rest. Of patients with valvular heart disease, ∼30% are in each of the classes I, II and III and ∼10% are in class IV. 1 Most patients with severe AS display symptoms of disease. However, this is not always the case and some patients with haemodynamically severe AS are asymptomatic. 7

Epidemiology

Aortic stenosis is the most common valvular heart disease in Western countries. 2,3,5 The majority of people treated for AS are > 60 years of age1 and the prevalence of AS has been found to be 2% in those ≥ 65 years, 3% in those ≥ 75 years8 and 4% in those ≥ 85 years. 9 The prevalence of valve disease is strongly linked to population ageing10 and the prevalence of AS is therefore likely to increase in developed countries such as the UK because of the increasing proportion of elderly people in those populations.

Natural history

The progression of AS is very variable, with some patients living a long time with stable, asymptomatic disease. However, once symptoms develop, risk of death is high. 2,11 Symptoms may develop in patients of any age but most commonly present in patients in their 60s. 11

In patients with severe symptomatic AS, in the absence of treatment, survival rates have been shown to display a sharp decrease, independent of age at symptom onset. 11 Mortality risk is about 2% per month4 and typical survival is < 2 to 3 years. 5,12 Two-year mortality is approximately 50%13 and 3-year mortality approximately 75%. 4 In a trial which enrolled patients unsuitable for surgical intervention, the 1-year mortality under standard therapy was approximately 50%. 14 This higher mortality is not unexpected given the patient group involved.

In asymptomatic patients, sudden death is rare2 but there is a small risk that symptoms will develop very rapidly and the patient will die suddenly or even that death will occur suddenly without the onset of clinical symptoms. The risk of sudden death in asymptomatic patients has been estimated to be around 1% per year. 2,7

Surgical aortic valve replacement

Medical management

Balloon aortic valvuloplasty

Transcatheter aortic valve implantation

Current treatment guidelines

In 2008, NICE published its initial guidance for the use of TAVI in patients with AS; this was updated in 2012. As well as defining the relevant patient groups (see Transcatheter aortic valve implantation, Patients), the following guidance was given:25

Clinicians wishing to undertake TAVI for patients with aortic stenosis for whom SAVR is considered suitable but to pose a high risk. . .should take the following actions.

1. Inform the clinical governance leads in their Trusts.

2. Ensure that patients understand the risk of stroke and death, and the uncertainty about the procedure's efficacy in the long term. Provide them with clear written information. In addition, the use of NICE's information for patients (‘Understanding NICE guidance’) is recommended.

Patient selection should be carried out by a multidisciplinary team including interventional cardiologists, cardiac surgeons, a cardiac anaesthetist and an expert in cardiac imaging. The multidisciplinary team should determine the risk level for each patient.

TAVI is a technically challenging procedure that should be performed only by clinicians and teams with special training and experience in complex endovascular cardiac interventions. Units undertaking this procedure should have both cardiac and vascular surgical support for emergency treatment of complications.

NICE encourages further research into TAVI for aortic stenosis. In particular, NICE encourages clinicians to enter all suitable patients into the UK TAVI trial. Information from research trials that will be useful for future guidance includes patient selection criteria and comparisons between TAVI and SAVR in patients who would be suitable for either procedure. Outcomes should include incidence of stroke and other adverse events, symptom relief, quality of life, occurrence of aortic regurgitation, and valve durability in the short and long term.

Other treatment-related guidance has been issued by the European Society of Cardiology (ESC), European Association of Cardio-Thoracic Surgery and the European Association of Percutaneous Cardiovascular Interventions. Guidelines have been set out for the use of SAVR, balloon valvuloplasty and medical treatments2 and TAVI10 in patients with severe AS. The following is recommended:

For surgical aortic valve replacement:

• Early valve replacement in all symptomatic patients with severe AS who are otherwise candidates for surgery.

• Early elective surgery, at the asymptomatic stage can only be recommended in selected patients, at low operative risk.

For balloon valvuloplasty:

• Balloon valvuloplasty can be considered as a bridge to surgery in haemodynamically unstable patients in whom surgery is associated with high risk or in patients with symptomatic severe AS who require urgent major non-cardiac surgery.

• Occasionally, balloon valvuloplasty could be considered as a palliative measure in individual cases when surgery is contraindicated because of severe comorbidities.

For medical management:

• Patients who are unsuitable candidates for surgery may be treated with digitalis, diuretics, ACE inhibitors or angiotensin receptor blockers if they are experiencing heart failure.

For transcatheter aortic valve implantation:

• Patient selection should involve a multidisciplinary consultation between cardiologists, surgeons, imaging specialists, anaesthesiologists and possibly other specialists if necessary.

• TAVI is indicated in patients with pure or predominant AS.

• TAVI should be performed only in patients with severe AS.

• At present, TAVI should be proposed only for patients with severe symptoms that can definitely be attributed to valve disease.

• TAVI should currently be restricted to patients at high risk or in whom surgery is contraindicated.

• It is premature to consider using TAVI in patients who are good surgical candidates.

• TAVI should not be performed in patients whose life expectancy is < 1 year, and who should be managed conservatively.

Current policy and demand

In 2005–6 there were 10,396 hospital diagnoses of AS in England. 35 If it is assumed that this incidence is typical and that most of these cases will, at some point, go on to present as severe AS, it can be estimated that around 10,000 people are assessed for severe AS each year.

A position statement, published in November 2010 by the British Cardiovascular Intervention Society (BCIS) and the Society for Cardiothoracic Surgery (SCTS),36 states that ‘TAVI should currently be reserved for patients who have been considered by a multidisciplinary team (including 2 surgeons and 2 interventional cardiologists) who consider the risk/benefit ratio of open heart surgery and TAVI to favour TAVI’.

A large number of patients with severe AS may be at low procedural risk and be referred directly for SAVR. However, for patients considered to be high risk for SAVR, the projected use of TAVI is unclear. For this latter patient population, the balance of treatment with SAVR and TAVI may depend on many factors (see Choice of treatments), some of which will change with time. The number of TAVI procedures in the UK has increased, from 67 people in 2007 and 272 in 2008, to 533 in 2009 (P Ludman, UK TAVI Registry presentation on behalf of UK TAVI registry group, 2010, personal communication). TAVI is currently used relatively infrequently and it appears likely that, although the exact uptake is unclear, the number of TAVI procedures in the UK is likely to rise over time.

Patient pathway

The patient pathway is likely to vary between centres and may depend on staff resources and the length of time that centres have been undertaking SAVRs or TAVIs. Referral may often take the following course. Patients with pre-diagnosed AS may be under the watch of their general practitioner (GP). On the development of symptomatic disease that is judged to be severe, or on the initial presentation of a patient with severe symptomatic AS, patients may be referred to a cardiologist for further/initial investigation and assessment. If the cardiologist believes further intervention to be appropriate, patients are referred to a multidisciplinary team (MDT) to discuss treatment options. The MDT usually consists of cardiology and surgical specialists and an anaesthetist, with expertise and experience in the treatment of AS and TAVI/SAVR techniques. Together, this team discusses each patient individually and comes to a consensus decision about the best course of treatment.

Where patients are deemed eligible for conventional SAVR, they are referred for surgical assessment. Where patients are considered too high risk for SAVR, they may be considered for medical management or, if they are judged as potential candidates for TAVI, referred to a TAVI clinic. Here, there is further assessment to determine the suitability of TAVI and, where appropriate, the best mode of entry. In some cases, where further discussion has deemed that a patient is fit to undergo surgery, these patients may be referred back for SAVR. Where doctor assessment or patient preference deems that TAVI is not appropriate or desirable, patients may be referred for medical management. Patients may be placed under review by the TAVI clinic and it is possible that balloon aortic valvuloplasty may be used at this point as a proof of concept that TAVI may be undertaken successfully.

There may be considerable fluidity in the referral process between medical and surgical teams, with patients originally assigned to one intervention (medical management, TAVI or surgery) subsequently being referred for another. Patient preference will also play a role in the referral process and may ultimately be the factor that determines which treatment is undertaken; for example, a patient offered TAVI might decline it and accept medical management instead. Figure 1 outlines an example of a possible referral pathway for patients with severe AS.

FIGURE 1.

Example patient referral pathway for patients with severe AS.

Choice of treatments

A variety of factors are taken into account in the decision-making, and the choice between SAVR, TAVI and medical treatment depends on the relative risks associated with procedures/treatments, the potential gains in terms of extended lifespan and QoL, and the preferences of the patient and medical team. Factors governing these decisions are likely to include risk scores, comorbidities, patient age, patient preference, and the expertise and preference of medical staff.

Referral for surgery

The most appropriate course of treatment is often unclear, and there is likely to be some degree of subjectivity in the decision-making process. For example, in the US PARTNER trial of patients assessed to be inoperable and randomised to medical therapy, 10% went on to undergo SAVR in the following year. 14 In another study, of patients who had not initially been proposed for surgery due to comorbidities, 75% subsequently went on to have SAVR in the following months. 52 Decisions may be influenced by the opinions and preferences of the assessing physician or surgeon. A limiting factor for SAVR may be the referral of patients for surgical assessment. In a study of patients with severe AS, 69% were not referred for surgical assessment and did not undergo SAVR. 53 All of the other patients, who were assessed by cardiothoracic surgeons, went on to have SAVR,53 suggesting that surgical referral was the limiting factor in determining patient treatment.

With the development of the TAVI procedure, it is now recommended that MDTs assess each patient to determine their suitability for SAVR, TAVI or medical management. 25 With the increasing use of MDTs for patient assessment, surgical referral may no longer be a limiting factor for SAVR and the results of this have been observed. When comparing rates between the years before and after the introduction of TAVI, US investigators noted a 64–74% increase in rates of surgical referral and a 44% increase in SAVRs, whereas the number of un-operated AS patients fell from 51% to 41%. 51

The impact of TAVI on the number of SAVRs in a UK centre has also been studied. 54 At the University Hospital of South Manchester (UHSM), in the 2 years after the introduction of the TAVI service, there was a 37% increase in the numbers of SAVR procedures, as shown in Figure 2.

FIGURE 2.

Surgical aortic valve replacement activities at the UHSM from January 2005 to December 2009. Redrawn from Grant et al. 54

Study investigators concluded that this increase was largely due to the increased awareness of treatment options and referral for SAVR. The authors also highlight the importance of the MDT in this process. Patients are considered first for SAVR, before being referred for TAVI. However, in practice, some patients are initially considered ineligible for SAVR and referred for TAVI, but after reassessment by the MDT may subsequently receive SAVR. 54 This has been demonstrated in other series of patients referred for TAVI, of whom 15%55 or 20%56 went on to undergo SAVR.

The increased use of MDTs in the setting of AS treatment decision-making may have had a beneficial impact in providing balanced, multidisciplinary assessments. It has been suggested that surgeons be the gatekeepers of TAVI, assessing the risk of standard SAVR and referring for TAVI if they believe the risks of conventional surgery to be too high. 57 However, assessment by MDTs with both surgical and medical expertise may more effectively bring to light all possible issues for selecting the best mode of treatment.

Post-transcatheter aortic valve implantation/surgical treatment and surveillance

Immediately after TAVI has been conducted, aortography and, when available, transoesophageal echocardiography or transthoracic echocardiography are performed to assess whether or not there is aortic regurgitation, to assess the patency of the coronary arteries, and to rule out complications such as haemopericardium and aortic dissection. 10 Haemodynamic measurements are also assessed using pressure recordings and/or echocardiography. 10

It is recommended that, following TAVI, patients should stay in intensive care for at least 24 hours and be closely monitored for several days, especially with regard to haemodynamics, vascular access, rhythm disturbances and renal function. 10 UK data show that the mean duration of hospital stay following SAVR is 10–11 days. 58 In the case of elderly patients (> 80 years), the average length of postprocedural hospital stay is around 13 days. 16

After discharge, patients may be referred for a programme of rehabilitation. This consists of exercise sessions aimed at reducing symptoms associated with AS and improving physical performance. 59,60 However, rehabilitation may not be routinely offered to patients who have undergone aortic valve replacement (Birmingham cardiac rehabilitation services, 2011, personal communication) and, where programmes are recommended, uptake may be low. 16

Selection of treatment

This part of the model is a schematic representation of the range of patient groups being compared. It takes the form of a decision tree, as illustrated in Figure 3. The first division in the tree represents the policy choice of whether or not TAVI is available. The whole patient group is taken to follow each of these choices in turn, but then subdivided at later nodes. It is convenient for modelling purposes to divide first by the underlying condition of the patient, even though this may not become known to any clinician until later (or, indeed, at all). Thus, patients are divided into those who are suitable for surgery and those who are unsuitable (too high risk and/or other reason such as suffering from porcelain aorta). The proportions following each of these branches must be the same for the two policy options (TAVI available or not available). Within these groups (suitable or unsuitable for surgery), patients are then divided according to the treatment actually selected. On clinical advice, and in the interests of simplicity, the possibility of balloon aortic valvuloplasty in the model has not been included.

FIGURE 3.

Selection of treatments for AS patients. The model structure takes the form of a decision tree. The first division of the tree represents the policy choice of whether or not TAVI is available. The next division is the true condition of the patient in terms of surgical risk. Patients then divide according to the treatment actually given. There are eight different pathways, labelled A to H. For example, pathway A represents the case where TAVI is available; the patient is suitable for, and receives, SAVR. Pathway B represents patients who are suitable for surgery, but receive TAVI. MM, medical management.

So, for example, in the absence of TAVI, patients are either suitable for SAVR or not. Those who are suitable for SAVR may receive it (group F) or may not for a reason other than clinical risk (group G), an example being patients who choose not to undergo such invasive surgery. This group (G) would still receive medical management in the absence of TAVI, as would those who are unsuitable for SAVR (group H).

If TAVI is available, patients suitable for SAVR could receive any of SAVR (group A), TAVI (group B) or – for similar reasons to those given above – medical management (group C). Those not suitable for SAVR could receive either TAVI (group D) or medical management (group E).

This grouping of potential patients allows for all possible policy options relating to the use of TAVI, including the possibility that TAVI could be offered as an alternative to SAVR in certain patients.

These populations can be envisaged diagrammatically in terms of SAVR risk. For example, Figure 4 shows the distribution of patients in the case where TAVI is not available (groups F to H). The area of each region approximates to the proportion of patients following each possible pathway. The horizontal scale represents schematically the proportion of patients in order of increasing surgical risk. It is stressed that this surgical risk is not intended to be simply the value of some risk score, but is intended to take account of clinical features and other personal characteristics. It is assumed that the probability of surgery being chosen decreases with increased operative risk and that surgery is never offered if the risk is above a threshold. The diagram is drawn to reflect the possibility that a patient at minimum surgical risk may still end up with medical management.

FIGURE 4.

Distribution of patients when TAVI is not available. This graph represents the possible clinical treatment options for AS patients in the absence of TAVI. The patients are distributed from left to right according to their increasing surgical operative risk, based on a risk score adjusted for individual patient characteristics. Patients suitable and unsuitable for surgery are differentiated by a cut-off value that is the maximum acceptable risk for surgery. Each area (labelled F to H) represents the approximate proportion of patients who follow the appropriate arm of the decision tree in Figure 3. MM, medical management.

The availability of TAVI can be incorporated into this diagrammatic framework. Figure 5 shows the distribution of patients when TAVI is available. Some patients suitable for SAVR may receive TAVI; this could include both patients who would receive SAVR in the absence of TAVI and some patients who choose not to undergo surgery. Some patients unsuitable for SAVR would receive TAVI rather than medical management. In the base-case analysis, this last group represents 90% of the patients receiving TAVI, a further 9% being those who would have received SAVR.

FIGURE 5.

Distribution of patients when TAVI is available subject to a minimum SAVR risk. This graph represents the possible clinical treatment options for AS patients when TAVI is available only for patients who are above a minimum surgical risk. The patients are distributed from left to right according to their increasing surgical operative risk, based on a risk score adjusted for individual patient characteristics. Patients suitable and unsuitable for surgery are differentiated by a cut-off value that is the maximum acceptable risk for surgery. Each area (labelled A to E) represents the approximate proportion of patients who follow the appropriate arm of the decision tree in Figure 3. MM, medical management.

Figure 5 is drawn on the assumption that there is a threshold SAVR risk below which TAVI would not be allowed. This may be considered as a middle case where TAVI is available. Two extreme cases are possible. Removing the threshold SAVR risk and allowing TAVI to be used without restriction is illustrated in Figure 6. The opposite extreme, where TAVI is only available for patients in whom SAVR is contraindicated, is shown in Figure 7.

FIGURE 6.

Distribution of patients when TAVI is available with no restriction. This graph represents the possible clinical treatment options for AS patients when TAVI is available for all of them. The patients are distributed from left to right according to their increasing surgical operative risk, based on a risk score adjusted for individual patient characteristics. Patients suitable and unsuitable for surgery are differentiated by a cut-off value that is the maximum acceptable risk for surgery. Each area (labelled A to E) represents the approximate proportion of patients who follow the appropriate arm of the decision tree in Figure 3. MM, medical management.

FIGURE 7.

Distribution of patients when TAVI is available only to patients in whom SAVR is contraindicated. This graph represents the possible clinical treatment options for AS patients when TAVI is available only to those for whom SAVR is contraindicated, and is shown as a theoretical extreme, which is unlikely to represent true practice. The patients are distributed from left to right according to their increasing surgical operative risk, based on a risk score adjusted for individual patient characteristics. Patients suitable and unsuitable for surgery are differentiated by a cut-off value that is the maximum acceptable risk for surgery. Each area (labelled A to E) represents the approximate proportion of patients who follow the appropriate arm of the decision tree in Figure 3. Note that compared with the distributions in Figure 5 and Figure 6, there is no group B in this diagram. MM, medical management.

Results of treatment

For each patient group, the results of treatment were calculated by estimating overall survival and survival free of hospitalisation. QoL scores were attached to the two possible health states hospital-free survival and other survival, the latter state including all surviving patients who had undergone at least one episode of hospitalisation after initial treatment. Monthly costs were attached to each health state, as well as a cost for first hospitalisation following initial treatment. For simplicity, further hospitalisation was not modelled explicitly, with the costs and QoL scores for the other survival state intended to represent an average of all patients in that state. As there is no explicit provision in the model for the effect of long-term adverse events on QoL, this is assumed to be included in the data on which the QoL calculations are based. Average monthly costs and QoL were accumulated to estimate the total expected costs and QALYs per patient, applying an annual discount rate of 3.5% to both costs and outcomes.

The modelling framework allows for separate estimates to be made for each of the patient groups described above and is shown in Figures 3–7. In practice, a single estimate was used for the two TAVI groups (B and D), and a (different) single estimate was used for the four medical management groups (C, E, G and H). However, it was possible to estimate the mortality separately for low and moderate-risk groups for SAVR, and this was used where appropriate to give separate estimates for the two SAVR groups (A and F).

Fitting survival curves

Data were available for various populations in terms of both overall survival and hospitalisation-free survival. In all cases, the follow-up time was considerably shorter than is required for the modelling. To extrapolate these curves forward in time, it is necessary to fit a parametric curve to the data. Visual inspection of survival curves shows a decreasing hazard with time. However, this cannot be sustained in the longer term, as the general population mortality risk increases with age. It is necessary to represent the long-term mortality that can be specifically attributed to AS using a function that can be approximated by a parametric form suitable for extrapolation. Without making any assumptions, the overall survival S(t) at time t can be written as:

where k is the 30-day survival; f(t) is a function to be determined, and is related to the long-term mortality specific to AS; and g(t) is the general population survival.

For TAVI and medical management, survival curves are available for both overall survival and hospitalisation-free survival for up to 2 years from the PARTNER trial (Leon et al. , figure 1). 14 Reading from the Kaplan–Meier curves, the values in Table 3 were obtained. The entries in the first three columns were taken directly from the source; the remaining columns show the result of the process used to infer the function f(t).

It is convenient to fit a Weibull distribution for f(t) taking the form:

where a and b are parameters to be determined. Here a is a dimensionless shape parameter: if a = 1 then the distribution reduces to the exponential distribution with a constant hazard. If a < 1 then the hazard decreases with time, while if a > 1 the hazard increases with time. For a fixed value of a, the mean survival time represented by f(t) is proportional to the scale parameter b.

Fitting a Weibull distribution to a given set of observed values can be done by transforming the equation into the form:

and estimating values of a and b to minimise the sum of the squares of the errors in the above equation. In line with the principle of parsimony in modelling, a common value of the shape parameter a is to be preferred whenever an assumption of proportional hazards can be considered at all reasonable. A reasonable fit can be obtained to the last two columns in Table 3 by using a = 0.550 in each case, with b = 116 for overall survival, and b = 60.9 for hospitalisation-free survival. The comparison between the observed and fitted values can be seen in Table 4.

Similar calculations were performed for the medical management group, again using the values from the PARTNER trial. 14 For surgical aortic valve replacement, other data sources had to be used. 42,67 Patients undergoing SAVR were divided into two groups: low risk (LES  < 10)68 and moderate risk (LES between 10 and 20). 42,67 Separate curves were fitted for these groups. Two data sources were available for moderate risk and these were combined.

Full details of the relevant calculations are in Appendix 3. The parameters for the various curves fitted are shown in Table 5. Although the 30-day survival is slightly higher for the moderate-risk group than for the low-risk group, the long-term survival is much higher in the low-risk group. This difference is of little importance in the modelling, as almost all of the results are based only on the moderate-risk group.

As an illustration of how these numbers are then used in the model, consider the population at 12 months, for which time the relevant general population survival is 0.963. For TAVI, we have

From these figures, a proportion 0.686 – 0.571 = 0.115 of the original population is alive but has had at least one repeat hospitalisation. Note that the modelled values differ slightly from the observed figures in Table 3 for the same reason as the fitted values in Table 4 differ from the observed values in that table.

For medical management, there is no additional mortality assumed in the first month and the calculation at 12 months is as follows:

From these figures, a proportion 0.524 − 0.297 = 0.227 of the original population is alive, but has had at least one repeat hospitalisation.

Quality-of-life scores

The model requires average QoL scores for survivors with or without repeat hospitalisation. The available information from the PARTNER trial14 does not include direct estimates of these QoL scores. These must therefore be estimated. This was done using the two-stage process described below.

First, the PARTNER trial14 reports results in the form of proportions in each of the four NYHA classes at 6 months and 1 year. QoL scores (utilities) were obtained by reading from figure 3 in the paper of Maliwa et al. 69 From these figures, the mean QoL scores for the TAVI group at each time could be obtained as shown in Table 6. These are weighted averages of the QoL scores, weighted by the proportion of TAVI patients in each class at the relevant time. These mean QoL scores are also weighted averages of the QoL scores for survivors with or without repeat hospitalisation, weighted by the number of TAVI patients in each category at the relevant time. The PARTNER trial14 has also reported these numbers, which are also shown in Table 6.

The second stage of the calculation involves solving a pair of linear simultaneous equations, to identify the QoL scores for survivors with or without repeat hospitalisation, using the overall average QoL scores at each time, and the proportions of patients in each category. The values obtained were 0.717 for survival without repeat hospitalisation and 0.579 for survival with repeat hospitalisation. It should be noted that these point estimates are subject to considerable uncertainty given the nature of the method used.

To estimate the QoL under medical management, the standard therapy arm of the PARTNER trial14 was used. When this was done using a similar process to that described above, the average QoL scores at 6 months and 1 year were estimated to be 0.545 and 0.585 respectively. However, the PARTNER trial14 reports a higher proportion of repeat hospitalisation in the standard therapy arm among survivors at 1 year than at 6 months. If there were a lower QoL in survivors with repeat hospitalisation than in survivors without repeat hospitalisation, the average QoL should be lower at 1 year than at 6 months. As this is not the case, it is not possible to obtain sensible estimates of QoL scores for the two states separately for the medical management arm of the model using the same method as for TAVI. Instead, a single figure of 0.565 was found by averaging the 6-month and 1-year results.

Although the methods used to obtain the QoL scores are far from ideal, the results obtained seem reasonable. For patients undergoing SAVR, the same values were used as for TAVI in the absence of any alternative data source. However, in the first month of the model, account is taken of the time in hospital and post-surgery recovery. For SAVR, QALYs accumulated were reduced by 50%, whereas for TAVI they were reduced by 25% to reflect the lesser invasiveness of the procedure.

Costs

The costing perspective was NHS, given that the vast majority of the patients are over retirement age; the price year was 2010. 70 Costs in the model are detailed below. They are made up of the short-term costs associated with the procedures, and then longer-term costs resulting from follow-up appointments, medication and repeat hospitalisation. All costs and QALYs were discounted at 3.5% from the starting point.

Varying the comparator with transcatheter aortic valve implantation for patients suitable for surgical aortic valve replacement

In the base-case analysis, it was assumed that 10% of TAVI cases would be patients suitable for SAVR, and that 90% of these would actually receive SAVR if TAVI were not available. Suppose that, instead, only 50% of these patients would receive SAVR. Compared with the base case, this scenario reflects what would happen if patients who receive TAVI but would have been suitable for SAVR are drawn in higher numbers from those who would not actually have received SAVR, for example due to personal choice not to undergo SAVR. The overall results and results for the suitable for SAVR group are then as shown in Tables 12 and 13. Results for the unsuitable for SAVR group are not shown, as they are unchanged from the base-case results.

Figure 9 shows the results of varying the time horizon in this case. At all time horizons, TAVI is now more effective, but still more costly, than the comparator, and so a positive ICER is shown. As in the base-case results, the overall results are similar to the results for patients unsuitable for SAVR.

FIGURE 9.

Varying the time horizon when only 50% of TAVI patients suitable for SAVR would actually have received SAVR.

An extension of this sensitivity analysis is to consider the patients suitable for SAVR and vary the proportion of them who would actually receive SAVR across the whole range from 0% to 100%. The effect of this is shown in Figure 10. This is based on a 25-year time horizon. The results for the unsuitable for SAVR group do not vary, whereas the results for the overall group remain close to those results throughout. The results for the suitable for SAVR group vary considerably according to the proportion who would actually have received SAVR. The ICER goes over £20,000 per QALY when this proportion reaches 36%, and over £30,000 per QALY at 50%. The ICER becomes very large as the proportion exceeds 67% and the TAVI option is dominated by the comparator when the proportion is 68% or higher.

FIGURE 10.

Varying the percentage of suitable patients receiving SAVR in the absence of TAVI.

Varying the eligibility criteria for transcatheter aortic valve implantation

This analysis relates to Figures 6 and 7 in Model description, where the eligibility for TAVI is either extended to all AS patients or restricted to patients who are deemed completely unsuitable for surgery. In fact there is no need for any further analysis in the latter case, as it is covered by the results in Table 11, with an ICER of £12,900 per QALY. If TAVI is extended to a wider range of patients with moderate risk, then the results for the two patient groups separately are still as in Table 10 and Table 11, but the overall results change. Table 14 shows the results in the case where equal numbers of TAVI patients are drawn from the two patient groups suitable and unsuitable for SAVR, assuming that 90% of TAVI patients would receive SAVR in the absence of TAVI. The overall ICER increases from its base-case value of £13,700 per QALY to £25,400 per QALY. The overall ICER crosses £20,000 per QALY when the proportion suitable for SAVR is around 40%, and £30,000 per QALY at around 55%. Finally, if as many as 76% of TAVI cases came from the group of patients suitable for SAVR, the net effect of introducing TAVI would be a reduction in overall QALYs at an increased overall cost.

The above analysis assumes that all patients receiving TAVI instead of SAVR would be taken from the moderate-risk group. If low-risk patients were also included, the results for TAVI would be worse. For example, if TAVI patients were drawn in equal numbers from those suitable and unsuitable for SAVR, and those suitable for surgery were drawn in equal numbers from low- and moderate-risk patients, then the overall ICER would increase to nearly £48,000 per QALY (Table 15).

Considering the effect of increased referrals

In the background section of this report, it was noted that a possible effect of TAVI availability is that additional patients might be referred for investigation and end up receiving SAVR when they might have had medical management in the absence of TAVI. As a speculative analysis, suppose that the total number of patients receiving SAVR is unchanged as a result of the introduction of TAVI. In the base-case analysis, 90 patients receive TAVI when it is available but would receive SAVR when TAVI is not available (see Table 8). Therefore, to maintain the total numbers receiving SAVR would involve including in the modelled population an additional 90 patients receiving SAVR in the TAVI available strategy and medical management in the TAVI not available strategy. Assuming that all of these patients are drawn from the group with moderate SAVR risk, the results both overall and for patients suitable for SAVR are shown in Table 16 and Table 17 respectively.

As can be seen, the results change dramatically and suggest that TAVI availability would be cost-effective in all groups. In fact, in this case, it can be seen that the ICER for TAVI availability is the same for patients suitable for SAVR as it is for patients not suitable for SAVR. This is not a coincidence: what is now happening is that TAVI is effectively displacing medical management only in both groups. Different people receive SAVR in the two arms of the model, but they are assumed to be taken from a homogeneous population, and so the change in identity of the SAVR patients has no effect on the model results.

The above analysis assumes that all SAVR patients are drawn from the moderate-risk group. If the additional SAVR patients are drawn in equal numbers from low- and moderate-risk groups, whereas the patients moving from SAVR to TAVI are still taken only from the moderate-risk group, the ICER for the patients suitable for SAVR drops even further to around £9800 per QALY.

It may well be the case that the number of patients receiving SAVR in place of medical management is considerably lower than the number receiving TAVI in place of SAVR. The ICER for TAVI availability in the group suitable for SAVR goes over £20,000 per QALY when approximately 60% of the patients switching from SAVR to TAVI are replaced by patients switching from medical management to SAVR, assuming again that all SAVR patients are moderate risk.

Summary of sensitivity analysis results for patient selection

Table 18 summarises the results presented in detail in the previous three sections. The results given here show that the overall ICER and the ICER for patients suitable for SAVR are highly sensitive to the rules for selection of patients for TAVI.

Allowing for the possibility of repeat transcatheter aortic valve implantation

As noted in the background section, the long-term durability of TAVI valves is currently unknown. The possibility that replacement TAVI may be necessary was considered by adding the cost of an additional TAVI to all surviving patients at a given time from the start of the model. On clinical advice, the times chosen were 7, 10 and 15 years. It is acknowledged that this is a very simplistic approach, but it at least gives some indication of the scale of the potential problem. The results are shown in Figures 11, 12 and 13. In each case, the part of the graph for a time horizon up to the point of valve replacement is the same as for the base-case result shown in Figure 8. There is then a sudden increase in the ICER as the cost of replacement TAVI is included in the analysis. The overall results are not greatly affected by this change to the model. Indeed, in the case of replacement after 15 years the effect of the change is hardly noticeable on the graph. This reflects an overall modelled survival rate of 7% at that time, together with the effect of discounting. In all cases, the TAVI available strategy is dominated by the TAVI not available strategy for patients suitable for SAVR, regardless of the time horizon of the model.

FIGURE 11.

Including cost of replacement TAVI for all patients after 7 years. a, ICER not shown for patients suitable for SAVR as TAVI is dominated by comparator for all time horizons.

FIGURE 12.

Including cost of replacement TAVI for all patients after 10 years. a, ICER not shown for patients suitable for SAVR as TAVI is dominated by comparator for all time horizons.

FIGURE 13.

Including cost of replacement TAVI for all patients after 15 years. a, ICER not shown for patients suitable for SAVR as TAVI is dominated by comparator for all time horizons.

Resource use in first month

There is considerable uncertainty about a number of the inputs from which the first month's costs for SAVR and TAVI were estimated. For example, in the base-case analysis, an estimated mean length of stay in hospital following SAVR of 13.36 days was used, made up of 3.1 days in an intensive care unit (ICU) and 10.26 days on a cardiology ward. Data from the Adult Cardiac Surgery Database58 give a mean stay of 10.7 days. If we use this figure for the total length of stay and keep the assumption of 3.1 days in an ICU, this reduces the mean days on a cardiology ward by 2.66 days, reducing the cost of SAVR by 2.66 × £396.16 = £1053.79. This clearly has no effect on the results for the patients not suitable for SAVR. For the patients suitable for SAVR, the average cost per patient for the TAVI not available strategy is reduced. Since this strategy was already less costly and more effective than the TAVI available strategy in our base-case analysis, the same must be true with the change in assumption. There is a small effect on the overall ICER, but the rounded value remains at £13,900 per QALY.

In the same way, any change in the assumption about frequency of short-term adverse events or inclusion of rehabilitation costs would have the same effect as an adjustment in the unit cost of the relevant treatment. Table 19 and Table 20 show the effects of changes in the unit cost assigned to SAVR and TAVI respectively.

In all cases explored, the TAVI not available strategy is less costly and more effective than TAVI available for patients suitable for SAVR. The results in Table 19 necessarily show no change in the ICER for patients not suitable for SAVR and only small changes in the overall results. The variation shown in Table 20 is somewhat greater, since the change in TAVI cost affects a larger number of patients in the model. However, the results still do not change greatly for fairly substantial changes in the unit cost assigned to TAVI.

Cost of repeat hospitalisation

In the base-case analysis, the cost of repeat hospitalisation was assumed to be approximately £4400 for SAVR and TAVI patients, but £7600 for medical management patients. If the lower figure is also used for medical management patients, the ICER for patients unsuitable for SAVR increases from £12,900 per QALY to £13,600 per QALY, and overall from £13,900 per QALY to £14,700 per QALY. Similarly, if the higher figure is used for all patients, these ICERs change to £13,300 and £14,300 respectively. In either case, the TAVI available option is still dominated for patients suitable for SAVR.

Short-term mortality

In this analysis, the short-term mortality was adjusted for each of SAVR and TAVI separately. For SAVR, the 1-month mortality was given as 7.3%, with a 95% CI from 4.3% to 11.0%. These results were obtained from combining the results of studies by Leontyev67 (10.7%) and Wendt42 (4.6%). For TAVI, the 1-month mortality was given as 5.0%, with a 95% CI from 2.3% to 8.7%. 14 Applying each of these figures individually gives the results in Table 21. They show very little variation from the base case.

Long-term mortality

The methods used to estimate the long-term survival curves do not naturally produce CIs. For deterministic analysis, an arbitrary decision has been taken to apply multipliers to shape and scale parameters of the Weibull distributions separately. Multiplying the scale parameter by a constant effectively multiplies the mean of the underlying distribution by that same constant; the effect of multiplying the scale parameter is not easily described. Shape parameters are naturally on a more limited range than scale parameters. Accordingly, multipliers of 0.8 and 1.2 have been used for shape parameters, while multipliers of 0.6 and 1.4 have been used for scale parameters. These correspond roughly to the 95% limits used in the probabilistic sensitivity analysis shown in Probabilistic sensitivity analysis. The results are in Table 22.

These results are not highly sensitive to the changes in the shape parameter, but are more sensitive to the changes made in the scale parameter. In particular, if the overall survival (relative to population survival) following SAVR is reduced by 40%, it is no longer the case that the TAVI not available strategy is more effective than the TAVI available strategy for patients suitable for SAVR. In this single scenario, TAVI available is both more costly and more effective than the comparator, but still well above the upper NICE threshold of £30,000 per QALY.

Quality-of-life scores

The deterministic sensitivity analysis for QoL scores is based on the idea of using 95% limits of an appropriate distribution to represent the uncertainty. There is no theoretically correct distribution for these parameters and it is necessary to make a pragmatic decision. It is reasonable to suppose that all QoL scores used in the model represent states better than death, and so the QoL scores must be between 0 and 1. Further, it is reasonable to suppose that there is a very low probability that the scores will be at either extreme of possibility.

A convenient distribution for QoL scores is a beta distribution. This has two parameters a and b: the mean of the distribution is equal to a/(a + b) and (for a fixed mean), the variance reduces as a and b increase. Ensuring that there is a very low probability of taking extreme values requires that both a and b parameters are at least equal to two. The approach taken here is to allow the maximum possible variance subject to this condition. There are three QoL parameters in the model:

u₁ is the QoL for hospitalisation-free survival following SAVR or TAVI;

u₂ is the QoL for post-hospitalisation survival following SAVR or TAVI; and

u₃ is the QoL for survival following medical management.

Consider first the QoL u₁ following surgery (SAVR or TAVI), but without repeat hospitalisation, for which the point estimate is 0.717. Since 0.717 > 0.5, we will have a > b, so it is appropriate to take b = 2 and then calculate a from the equation

giving the answer a = 5.06. This distribution has 95% limits 0.363 to 0.957.

For post-hospitalisation QoL after SAVR or TAVI u₂ it is important to keep the value lower than the QoL score u₁ for hospitalisation-free survival. It is therefore not appropriate to sample u₂ directly from a beta distribution. Instead, it makes sense to use a beta distribution for the ratio between the two QoL scores. The point estimate for this is given by

so the ratio can be taken to follow a beta(8.37,2) distribution, which has mean 0.807 as required. The 95% limits of this beta distribution are 0.532 and 0.973. These correspond to QoL scores u₂ of 0.532 × 0.717 = 0.381 and 0.973 × 0.717 = 0.697 respectively when u₁ takes its base-case value of 0.717.

Similarly, it is reasonable to suppose that the QoL for medical management u₃ should be lower than u₁. However, there is no compelling reason why u₃ should be either higher or lower than u₂. Accordingly, it is sensible to apply a beta distribution to the ratio of u₃ to u₁, in the same way as for u₂. This time the point estimate for the ratio is 0.788, which is the mean of a beta(7.44,2) distribution. The 95% limits for this distribution are 0.494 and 0.970, giving 95% limits for u₃ of 0.354 and 0.695.

For the univariate analysis shown in Table 23, each of the QoL scores was taken to its lower and upper 95% limits. When u₁ was varied, the other two scores were adjusted in proportion, but when u₂ and u₃ were varied, each of the other scores was kept at its base-case value.

The results show that the overall ICER and the ICER for patients not suitable for SAVR varied somewhat but remained below £20,000 per QALY in all cases, except the one in which the QoL scores were reduced to very low – and perhaps unrealistically low – values. In all cases, the option TAVI available was dominated for patients suitable for SAVR.

Summary of sensitivity analysis for parameters related to results of treatment

In all but two of the cases considered, the overall ICER for TAVI and the ICER for TAVI applied to patients not suitable for SAVR remained below £20,000 per QALY, while the TAVI available option was more costly and less effective than the comparator for patients suitable for SAVR. The only case in which the ICERs for TAVI (overall or for patients not suitable for SAVR) rose above £20,000 per QALY was one in which the QoL scores were reduced to extremely, and almost certainly unrealistically, low values. The only case in which TAVI available was not less effective than the comparator was when the long-term survival following SAVR was reduced substantially.

While the univariate sensitivity analysis shown in this section gives an idea of the relative importance of each parameter in the model, it will usually underestimate the overall uncertainty. The next section of this report describes an analysis in which multiple parameters are varied simultaneously.

The Cochrane Library (Wiley) 2010 Issue 4 (NHS Economic Evaluation Database)

Database: MEDLINE (Ovid) 1950 to week 3, November 2010

Database: EMBASE (Ovid) 1980 to week 46, 2010

Database: MEDLINE (Ovid) 1950 to week 3, November 2010

Database: MEDLINE (Ovid) 1950 to week 3, November 2010

Database: EMBASE (Ovid) 1980 to week 46, 2010

Database: MEDLINE (Ovid) 1950 to week 2, October 2010

Database: MEDLINE (Ovid) In-Process & Other Non-Indexed Citations 24 November 2010

Database: MEDLINE (Ovid) 1950 to week 3, November 2010

Database: EMBASE (Ovid) 1980 to week 47, 2010

The Cochrane Library (Wiley) 2010 Issue 4 (Cochrane Central Register of Controlled Trials, Database of Abstracts of Reviews of Effects, Health Technology Assessment) Issue11 (Cochrane Database of Systematic Reviews)

Database: MEDLINE (Ovid) 1948 to week 1, January 2011

Database: MEDLINE (Ovid) 1948 to week 4, January 2011

Database: MEDLINE (Ovid) 1950 to week 4, October 2010

Aortic stenosis

Aortic stenosis is a degenerative condition in which the aortic valve becomes progressively narrowed leading to gradual obstruction of left ventricular outflow at the level of the aortic valve. The prevalence of this pathology increases with age, the majority of people treated being above 60 years of age. 1 Chronic aortic stenosis causes pressure overloading of the left ventricle which becomes hypertrophied to maintain the stroke volume and cardiac output. Under these circumstances the patient may remain asymptomatic for many years. However with longstanding hypertrophy the ventricle will eventually become less compliant and symptoms of breathlessness, angina or collapse will develop. In the late stages of aortic stenosis the left ventricle will dilate. For patients who present in congestive heart failure mean survival is less than a year.

Different outcome measures are used to assess the severity of the condition and these are based on clinical assessment. They include the following:

• The New York Heart Association (NYHA) heart failure classification is used to classify the severity of breathlessness: from class I, in which the patient has no limitation in daily physical activity, to class IV, in which the patient is breathless at rest. NYHA is a functional classification system linking the patient's symptoms and quality of life to normal life qualities (www.americanheart.org).

• Haemodynamic assessment may involve measurement by echocardiography and/or Doppler. Aortic valve area (cm²) is assessed relative to body surface area (in m²). Aortic valve area < 0.6 cm²/m² indicates severe aortic stenosis.

• Transaortic gradient (mmHg) measures the blood volume flow rate through the aortic valve. Peak transaortic valve gradient > 64 mmHg and mean transaortic valve gradient > 40 mmHg indicates severe aortic stenosis. 2

Treatments for aortic stenosis (in the absence of TAVI)

Surgical aortic valve replacement is the reference treatment for aortic stenosis with around 60,000 operations conducted in Europe annually. 1 Surgical aortic valve replacement involves replacing the diseased valve with a prosthetic mechanical or biological valve through a median sternotomy and using cardiopulmonary bypass. This surgical procedure consists of an incision along the sternum, after which the sternum itself is divided to provide access to the heart and lungs for surgery. The cardiopulmonary bypass is a circulatory support technique that temporarily takes over the function of the heart and lungs during surgery, maintaining the circulation of blood and oxygen delivery to the body's tissues. 2

Surgical risk of existing treatments

Surgical aortic valve replacement carries a very high risk for some patients, particularly those who are elderly and/or who suffer from concomitant illnesses. Risk factors include age over 80, previous cardiac surgery, chronic obstructive airways disease, peripheral vascular disease, previous stroke with residual deficit, poor left ventricular function, renal failure, diabetes and hypertension. 3

Surgical risk assessment tools

• The European System for Cardiac Operative Risk Evaluation (EuroSCORE) calculates the predictive operative mortality of patients that undergo cardiac surgery. For high-risk patients, the more accurate logistic EuroSCORE is used. This model is most commonly used in Europe; this type of scoring system also considers particular combinations of the risk factors (www.euroscore.org). EuroSCORE is limited in risk stratifying patients for aortic valve replacement as it is essentially designed to risk stratify patients undergoing CABG. However it is the most widely used system in the UK. It is also straightforward to use – by completing an online scoresheet with various patient factors the logistic EuroSCORE will be provided for you in the form of a predicted mortality percentage.

• The STS score is a risk model developed by the Society of Thoracic Surgeons based on clinical and demographic data in an adult population and used to predict operative mortality and morbidity after cardiac surgery. This scoring system has been developed in USA. The high surgical risk is defined by an STS risk score of 10% or higher (on a scale of 0% to 100%, with higher scores indicating greater surgical risk). The model is based on clinical and epidemiological data of a given population who have received cardiac surgery (www.sts.org). This is quite a sophisticated model that gives not only mortality risk but also risk of major morbidity such as prolonged ventilation, stroke and renal failure

• The Ambler Risk score has been developed in UK and is able to predict the risk of in-hospital mortality for patients undergoing heart valve surgery. This scoring system seems to be the more accurate on target population considered in this report. In real practice surgeons evaluate the surgery mortality risk according to more than one risk-scoring system and different associated clinical conditions. (http://www.ucl.ac.uk/statistics/research/riskmodel/index.html)

The British Cardiovascular Intervention Society (BCIS) and the Society for Cardiothoracic Surgery (SCTS) state that the patient's operative risk should be assessed by a multidisciplinary team (MDT). The multidisciplinary team should comprise two cardiac surgeons, two interventional cardiologists, an imaging specialist, cardiothoracic anaesthetists and experienced nurses who assess the cost/benefit ratio of open heart surgery and TAVI. The usual “High risk” patient eligible for TAVI will have a logistic EuroSCORE of ≥ 20 or an STS score of ≥ 10. 2

Interventions which do not involve open surgery

Percutaneous balloon valvuloplasty has a limited role in the treatment of severe aortic stenosis in adult patients. It may be considered to provide palliative treatment for patients with considerable co-morbidity as relief of symptoms is likely to be temporary. 2

Transcatheter aortic valve implantation

Transcatheter Aortic Valve Implantation (TAVI) is a procedure used as an alternative to open heart surgery for people with severe aortic stenosis. In highly developed nations, the request for valve surgery is increasing among older people, who may present with more co-morbidities and a higher incidence of concomitant coronary artery disease. TAVI may allow aortic valve replacement to be undertaken in some of these patients who would previously have been considered too high risk for aortic valve replacement. This is highly specialised technology and is relatively new: the first human case was in 2002. There is a UK TAVI registry; the number of TAVI per year included in this registry was 67 in 2007, 272 in 2008 and 533 in 2009. 4 TAVI is currently restricted to high-risk patients with severe aortic stenosis and absolute contraindications for surgery. 5

TAVI aims to implant a bioprosthetic aortic valve at the site of the native aortic valve through a percutaneous route. The choice of the vascular route has been developed in order to reduce surgical trauma and the use of cardiopulmonary bypass associated with valve replacement. During the procedure, a biological valve is crimped into a delivery catheter. The delivery catheter is inserted either in the femoral artery through a small incision at the top of the leg (known as transfemoral, percutaneous, endovascular and transluminal approach) or between the ribs through the apex of the heart (known as transapical or transventricular approach). The valve is guided to the heart using radiological visual guidance. Usually the route of choice is the transfemoral as this is deemed to be the least invasive for the patient. The transapical route is used if the transfemoral route is limited by atherosceloris or small calibre. 2

A balloon catheter is advanced via the arterial system or the left ventricle over a guide wire and positioned within the opening of the aortic valve. The existing aortic valve is dilated in order to make room for the prosthetic valve. The new valve, mounted on a metal stent, is manipulated into position and is either self expanding or deployed using balloon inflation. Deployment leads to obliteration of the existing aortic valve.

TAVI may be carried out under general anaesthesia or spinal aneasthesia. The procedure requires a combination of echocardiography and fluoroscopic imaging to ensure accurate deployment of the valve. Prophylactic antibiotics and anticoagulation medication are administered before and during the procedure. 2

Surgeons and cardiologists involved in TAVI undergo dedicated training in both patient assessment as well as the clinical procedure. In addition the valve companies provide a high level of technical support and proctors with considerable experience until the new centre has enough experience to run an independent programme.

Cost implications

TAVI is an expensive procedure, requiring a multidisciplinary team and substantial equipment to be available during the procedure. The approximate cost of the procedure has been estimated at £18,000. 1 However, successful implantation may reduce the need for later hospitalisation, so there is likely to be some cost saving to offset against the cost of the procedure.

Cost effectiveness review

Initial scoping searches have been carried out to assess the volume and type of literature relating to TAVI for aortic stenosis.

National TAVI and aortic valve replacement registry is the most updated source of data. A trial published on the New England Journal of Medicine in October 20107 and other cohort studies summarised in Annals of Internal Medicine8 also in 2010 represent relevant updated evidence about short term effectiveness. Ongoing trials have been found into the TAVI clinical trials registers searches.

A model published by Bazian in May 20081 contains TAVI cost data, the model is not a cost-effectiveness analysis but it contains both procedure-related costs and costs that may be incurred during one year of follow up for both TAVI and medical therapy.

Bazian reported that they had found no published studies or models that had assessed the cost-effectiveness of TAVI. We will carry out an appropriate search to determine whether this is still the case. If any relevant studies or models are found, we will appraise them using standard criteria9. Although we do not expect to find any/many studies, if any robust evaluations are identified, we will use them to inform our model.

Economic model

An economic model will be designed to represent the pathway of patients with severe symptomatic aortic stenosis who are high risk or unfit for surgery compared to standard therapy.

The standard therapy comparator will differ depending on the type of patient. For patients at high risk, but not contraindicated for surgery, the most appropriate comparator may be surgery whereas, for patients contraindicated for surgery, other therapies will be more appropriate comparators. It is recognised that there is likely to be considerable areas of grey where the most appropriate treatment route for a particular patients is debatable. However, in order to make modelling for this project feasible, two separate patient groups will be considered. As was done in the PARTNER trial7, the two patient groups will be assumed to be distinct:

Group 1 – Patients who are at high risk but are not contraindicated for surgery. The comparator for this group will be surgery.

Group 2 – Patients who are contraindicated for surgery. The comparator for this group will be other forms of standard therapy (not surgery).

Before and during the development of the model, a steering group, with expertise in TAVI and surgical techniques and other possible comparators, will be consulted to give guidance to the technical team on clinical pathways, treatment strategies and other factors that may influence the structure and content of the model. This group will also give clinical guidance to inform the suitability of model parameters. This group is currently being formed under the direction of our clinical expert (SR).

As far as possible, the model will be populated with data derived from the literature. Data required for the model based economic evaluation may include:

Literature searches to obtain data

In order to obtain data to populate the model, our starting point will be the 2008 Bazian report1, the trial as reported in the NEJM,7 and the review of observational studies in the Annals of Internal Medicine. 8

We will update the searches in the Bazian report to ensure that up to date information is included. We consider these searches to be sufficiently robust. Wherever possible, model parameters will be chosen from studies where the data is most relevant to the decision problem.

Additional searches

Additional targeted searches may be performed on an ad hoc basis to seek information to populate parameters identified by the modellers that have not been obtained by other means.

Such information may include unit costs or prices, required to be attached to each resource item so that the overall cost per patient can be calculated. Some hospital resource utilisation and costs data are reported in the Bazian report1 but the two main sources will be the “Unit cost of health and social care10” published by (PSSRU) Personal Social Services Research Unit 2009 and the NHS Reference Cost. 11

Other sources of data

It is anticipated that data from controlled trials on long term outcomes and quality of life may not be found in the literature searches. Additionally, there does not appear to be data available from controlled trials of the effectiveness of TAVI in patients who are high risk, but not contraindicated, for surgery (patient group 2 in the model).

In order to address these points of concern, we have made contact with the study author of the PARTNER trial. Findings for one arm of this randomised controlled trial are currently published. 7 The published data is for patients who are contraindicated for surgery (patient group 1 in the model). The study author has provided us with data on patient quality of life at 1 year.

The other arm of the PARTNER trial involves the randomisation of patients who are high risk, but not contraindicated, for surgery (patient group 2 in the model). We have requested data from this arm of the trial. These results are due to be presented in spring 2011 and, if it is available before March 2011, the authors have provisionally agreed to provide us with that data.

A further source of data may be the national cardiac surgery register. We will take advice from the steering group whether this is a useful source of data for the model and, if so, how best to access the relevant information.