Base-case deterministic results: research question 1

For the base case we compared the cost-effectiveness of BTT based on all VAD recipients compared with the MM patients in the BTDB who were classified as ‘inotrope’. Derivation of utilities is reported in Chapter 4, transition probabilities between health states in Chapter 7, and resources and costs are reported in Chapter 8. Model inputs are summarised at the end of Chapter 8.

Results are tabulated (Table 68) in terms of mean cost, mean LYG and mean QALYs gained in each treatment group for 3-year, 10-year and lifetime (50-year) horizons. Also presented are the ICERs for these time horizons. The perspective is from the UK NHS and discounting of benefits and costs at 3.5% was undertaken according to UK guidelines.¹¹⁵

TABLE 68

Base-case deterministic results

For VAD patients compared with the inotrope subgroup of the MM patients the ICER is £122,730 per QALY over a 3-year time horizon. At the 10-year time horizon the ICER increases to £68,088 per QALY, and at a lifetime horizon of 50 years the ICER is £55,1730 per QALY.

The cost of the VAD and of the implantation procedure together make a substantial contribution to the costs. The impact of this and of other inputs is explored in sensitivity analysis. Undiscounted LYG in the BTT and the MM arms are summarised in Figures 61 and 62.

FIGURE 61

Undiscounted LYG with VAD – BTT.

FIGURE 62

Undiscounted LYG with MM support.

Base-case probabilistic results: research question 1

Base-case probabilistic inputs are shown in Table 67. The base-case probabilistic results summarised in Table 69 indicate a lifetime horizon ICER of £53,527 per QALY. Deterministic and probabilistic ICER estimates are similar for all three time horizons with the ICER increasing as the time horizon decreases.

TABLE 69

Base-case probabilistic results: (estimated using Monte Carlo method of 1000 simulations)

The joint distribution of the difference in costs and the differences in QALYs for the three time horizons is shown in Figure 63. Each point is a simulation from the joint distribution; the plot illustrates the uncertainty surrounding incremental costs and benefits for the two groups being compared.

FIGURE 63

Cost-effectiveness plane for 3-year, 10-year and lifetime horizons: base-case probabilistic results.

Figure 64 presents the CEACs for the 3-year, 10-year and 50-year time horizons. NICE guidance,¹⁰¹ although not explicit, suggests a benchmark of approximately £30,000 per QALY as the usual upper limit for the NHS. At a willingness to pay of £50,000 per QALY BTT approaches cost-effectiveness compared with MM (see Figure 63). Although, again not explicit, NICE appears to have applied a threshold of £50,000 per QALY for interventions which satisfy recommended criteria as end of life treatments. These include:

FIGURE 64

Cost-effectiveness acceptability curves for 3-year, 10-year and lifetime models.

• predicted survival of < 2 years in the absence of the intervention
• the intervention prolongs survival by at least 3 months
• a small population eligible for the treatment.

Sensitivity analyses: base-case analysis-research question 1

Sensitivity analyses were conducted by altering base-case inputs to the model. Several types of sensitivity analysis were explored encompassing changes to:

1. I] TPs between health states (I A to I D)
2. II] inputs for costs (II A to II D)
3. III] utility inputs for health states.

I] Impact of changing the transition probabilities between health states

The results for these sensitivity analyses are summarised in Table 70.

TABLE 70

Sensitivity analyses based on changes to TPs between health states

In analyses A2, the survival under MM was modelled according to the SHFM score after Levy et al.⁹⁵ for BTT patients. Using data from Schaffer et al.⁷⁷ and from Strueber et al.⁸³ the resulting modelled median survivals were 8.9 months and 16.5 months, respectively, providing values both higher and lower than that modelled for the BTDB inotrope patients (9.1 months). The resulting lifetime ICERs of £55,058 per QALY and £51,731 per QALY differed little from the deterministic base-case value of £55,173 per QALY. More recently Aaronson et al.⁹⁴ reported that the SHFM predicted 43% survival at 1 year for a MM group equivalent to the 140 BTT patients investigated in a HW study. This equates to a median predicted survival of 9.86 months, again very close to that modelled from our BTDB inotrope patients. Thus, when survival under MM is modelled according to these SHFM scores, the ICER estimates remain similar to that of the base case.

Some UK clinical experts expressed the view that a median survival of 9 months was too generous an estimate for inotrope-dependent MM patients entered onto UK lists for HT. On their suggestion sensitivity analysis was therefore undertaken in which the TP for MM to death was modelled on the survival of the optimum MM control group of the REMATCH trial (median survival 4.94 months); the resulting lifetime ICER (£55,203/QALY) was hardly different to that in the base case. The reason for this is that although the poorer survival of the MM arm results in an increase in the difference in QALYs between BTT and MM this poorer survival also results in lower costs in the MM group and an increase in the difference in costs between BTT and MM, and these factors tend to cancel out when calculating the ICER. It is interesting that under the base-case scenario, varying the survival of the MM arm between 3.9 and 16.5 months has negligible impact on the resulting lifetime ICER.

When the comparator population is constituted from the whole BTDB MM population (analysis A1), the ICER indicates that BTT is dominated, that is BTT is found to be more costly and less beneficial than MM. Similarly when a high probability of receiving a HT is applied to both groups (C1i), or if the MM arm is allocated a high probability but the BTT arm a low probability as in previous analyses (C1ii), BTT is dominated or the ICER becomes extremely large. These results indicate the critical importance of both the selection of an appropriate comparator population and of ensuring that an equal opportunity of receiving a HT is allocated to both groups.

These alternative scenarios have been modelled over the lifetime of the patient (i.e. until all patients have died). As in the base case, models with shorter time-horizons result in higher ICERs.

II] Impact of changing inputs for cost

Analysis II A change to device cost

We reduced the mean cost of the VAD by 10–76% to identify its impact on the ICER. The largest reduction in ICER was noticed at 76% reduction, where the 10-year and lifetime horizons of the model were more cost-effective (Table 71). Although the device has already been priced and marketed, this sensitivity analysis may inform reimbursement agencies in identifying the maximum price they may be willing to accept on the basis of cost-effectiveness.

TABLE 71

Incremental cost-effectiveness ratio based on reduced VAD cost (analysis II A): BTT vs. MM

These results indicate that under the base-case scenario a modest reduction in device cost of 15% reduces the ICER to a threshold of ∼ £50,000 per QALY, which is close to values used by NICE for treatments which satisfy end of life criteria. To bring the ICER to a threshold of £30,000 per QALY requires a very substantial reduction to the cost of the device of 76%.

Analysis II A: Change to patient maintenance cost supported on ventricular assist device

A sensitivity analysis considered that patients implanted with second- and third-generation VADs rather than first-generation devices experience fewer adverse events. The base-case monthly cost of immediate and long-term follow-up under VAD support was based on appropriately adjusted data from a previous study obtained in an era of transition between use of early and later generation devices. A reduction of 30% to this base-case cost resulted in an ICER of £42,914 over a lifetime time horizon (Table 72). It should be borne in mind that to date there are no firm data to support a conclusion that patients experience 30% fewer adverse events after implantation of second- and third-generation VADs. However, one publication identified in the clinical effectiveness systematic review (Ventura et al.⁸²) reported a non-randomised comparative study of HMII (n = 484) compared with the pulsatile HMXVE (n = 673) finding a significantly higher rate of hospitalisation for infection post implant for the pulsatile device. In addition, the RCT of DT with patients ineligible for HT conducted by Slaughter et al.,⁴⁷ comparing HMII with the pulsatile HMXVE device, reported lower risk in the HMII group for a wide range of adverse events (bleeding, stroke, rehospitalisation) and statistically lower rates of infection.

TABLE 72

Incremental cost-effectiveness ratio based on reduction in immediate and long-term monthly costs of VADs by 30%: comparison BTT vs. MM support

Analysis II C: Sensitivity analysis around costs for both arms using Golden Jubilee National Hospital data

Sensitivity analysis used variations on cost inputs for both arms based on the GJNH finance department costs; results are shown in Table 73. All ICERs (3 year, 10 year and lifetime) are higher than base case with these alternative costings.

TABLE 73

Incremental cost-effectiveness ratio based on health states cost sourced from the GJNH finance department 2010/11: comparison BTT vs. MM support

Analysis II D: Sensitivity analysis around costs for both arms using national schedule of reference costs data

In this sensitivity analysis we altered costs for both arms using variations in cost inputs based on the NSRC for 2010/11;¹²² results are shown in Table 74. All ICERs (3 year, 10 year and lifetime) are higher than base case with these alternative costings.

TABLE 74

Incremental cost-effectiveness ratio based on health states cost sourced from the NSRC: comparison BTT vs. MM support

III] Impact of changing utility values for health states

Univariate sensitivity analysis was undertaken replacing base-case utilities by those reported by Sharples et al.;³⁰ no large deviation in ICER was noticed (Table 75).

TABLE 75

Incremental cost-effectiveness ratio based on changes in utility score: comparison BTT vs. MM support

Tornado diagram

In sensitivity analysis the sources of all major base-case inputs were retained, but their values were raised and lowered at a fixed rate of 30% from their original values. For each parameter change, the percentage impact on the ICER is shown graphically in the form of a tornado diagram (Figure 65).

FIGURE 65

Tornado diagram. The bars indicate the effect on the base-case deterministic ICER (£55,173/QALY) of a 30% increase or decrease in input values for each of the input parameters listed on the right-hand side of the figure. Note: change to monthly (more...)

These analyses indicate that the most influential inputs were the monthly cost on BTT support, the monthly cost on MM support, and utility on VAD support. The probability of death while supported with a VAD was not influential in this over a ± 30% range of change. These results coincide with findings from the previous sensitivity analyses except that the time to HT is not influential in this analysis. This is because here the opportunity of receiving a donor heart has been kept the same for both VAD (BTT) and MM arms.

Base-case deterministic results: research question 2

The base-case TP inputs as listed for BTT VAD recipients for research question 1, and corresponding costs and utilities were applied to the comparator arm. For the ATT arm all inputs were the same as for the BTT arm except that the probability of receiving a donor heart was set to zero. It is recognised that ATT for patients suitable for HT is not currently a therapeutic option for the UK HF patients.

The base-case results for the deterministic model are represented for 3-year, 10-year and lifetime time horizons of the model, and these are tabulated in Tables 76 and 77. We have also presented results on a cost-effectiveness plane together with a CEAC (see Figures 66 and 67).

TABLE 76

Deterministic results for VAD-ATT compared against VAD-BTT

TABLE 77

Probabilistic results for VAD as ATT compared with VAD as BTT

FIGURE 66

Cost-effectiveness planes for 3-year, 10-year and lifetime time horizons. The dashed line indicates a saving of £30,000 for the sacrifice of 1 QALY.

FIGURE 67

Cost-effectiveness acceptability curve for VAD as an ATT vs. VAD as a BTT.

The ICERs (cost/QALY) for VAD as an ATT compared with VAD as a BTT over the 3-year, 10-year and lifetime study periods are £353,467, £31,685 and £20,637, respectively, but it should be noted that ATT costs less than BTT and delivers reduced benefit.

Over 3 years the ATT arm cost £10,604 less than the BTT arm and generated 0.03 fewer QALYs. At 10 years the ATT arm cost £15,329 less than the BTT arm and generated 0.48 fewer QALYs, and over a lifetime the VAD as ATT arm cost £32,813 less and generated 1.59 fewer QALYs. Thus, over a 50-year time horizon 1.59 QALYs are sacrificed at a cost saving rate of £20,637 per QALY.

Base-case probabilistic results: research question 2

Base-case probabilistic results are summarised in Table 77 and indicate a lifetime horizon ICER of £21,393 per QALY. The ICER falls mainly across the south-west quadrant, with a few results in the north-west quadrant, indicating that a VAD as an ATT is less effective and in some of the simulations is more costly. VADs as an ATT is, on the whole, cheaper – but confers less health gain. These findings are illustrated graphically in Figures 66 and 67. The cost-effectiveness plane for 3-year, 10-year and lifetime probabilistic estimates for VAD as an ATT compared with VAD as a BTT are shown in Figure 66 and base-case results are presented as CEACs for 3-year, 10-year and lifetime time horizons of the model in Figure 67.

These findings tell us that, relative to VAD for a BTT, VAD as an ATT over a 10-year or lifetime time horizon costs less and confers less benefit. At 10 years the intervention is just above ∼ £30,000 per QALY plane – albeit mainly within the ‘south-west’ rather than the ‘north-east’ quadrant.

In the final chapter we summarise our findings, discuss the strengths and limitations of the work and make recommendations for future research.