Clinical effectiveness and cost-effectiveness systematic reviews

Electronic databases were searched for relevant published literature on the clinical effectiveness and cost-effectiveness of CRT for heart failure. Studies comparing CRT and automatic ICDs alone were deemed outside the scope of this report. Updated searches were undertaken in June 2006. Included randomised controlled trials (RCTs) were critically appraised for internal and external validity. Relevant data were extracted, narrative reviews were undertaken and meta-analyses of the clinical trial data were conducted. The manufacturers' submissions to NICE were searched for additional evidence.

PenTAG cost–utility model

A Markov model was developed to address the NICE project scope and protocol. The model compared CRT-P and CRT-D with optimal pharmaceutical therapy (OPT) and with the alternative device, that is, CRT-P versus CRT-D. Ultimately, however, the results for CRT-D versus OPT were thought to be much less relevant to the decision problem, and the results from this comparison are accordingly given less attention. Clinical effectiveness parameters in the model were derived from the systematic review and other published sources. Resource use and costs associated with CRT and treating heart failure in the UK NHS were based on a mixture of published sources, unpublished sources and expert clinical advice. A simulated cohort of 1000 people of mixed age and mixed sex was modelled until the whole cohort was dead. Incremental costs and quality-adjusted life-years (QALYs) were calculated. Extensive one-way sensitivity analyses, threshold analyses, probabilistic sensitivity analyses and value of information analyses were carried out.

Number and quality of studies

Five RCTs met the inclusion criteria, recruiting a total of 3434 participants. Four studies compared CRT-P with OPT, two studies compared CRT-D with OPT and one study compared CRT-P with CRT-D. In all trials, patients with an indication for an ICD were excluded. Studies were of good to moderate quality. Two trials reported that allocation to treatment group had been concealed (CARE-HF and MIRACLE), blinding occurred in three trials (CONTAK-CD, MUSTIC-SR and MIRACLE) and intention-to-treat was used in four analyses (CARE-HF, COMPANION, MIRACLE and MUSTIC-SR).

Summary of benefits and risks

Meta-analyses showed that both CRT-P and CRT-D devices significantly reduced the mortality and level of heart failure hospitalisations. They also improved health-related quality of life in people with New York Heart Association (NYHA) class III and IV heart failure and evidence of dyssynchrony (QRS interval >120 ms) who were also receiving OPT. A single direct comparison (COMPANION) indicated that the effects of the CRT-P and CRT-D were similar, with the exception of an additional reduction in sudden cardiac death (SCD), associated with CRT-D. On average, implanting a CRT device in 13 people would result in the saving of one additional life over a 3-year period, compared with OPT.

There were no statistically significant subgroup effects, although trials were not individually powered for their detection.

A number of adverse events were associated with the CRT devices. The overall complication rate was 14%, mostly lead related. In addition, there were 0.8% perioperative deaths and 9% failure to implant the device. Of heart failure patients, 11–46% fail to benefit from CRT, clinical parameters suggesting a lower rate of failed response than echocardiographic measures.

Summary of costs

The NHS device and procedure cost of implanting a new CRT-P system (pulse generator unit and required leads) was estimated to be £5074, and that of a CRT-D system £17,266. Additional costs will be incurred for replacement devices at a mean time interval of 6.5 years for CRT-P and 5.5 years for CRT-D. The discounted lifetime costs of OPT, CRT-P and CRT-D (including the cost of periodic unit replacements, treatment for complications and potential implant with an ICD) were estimated as £9375, £20,804 and £32,689, respectively.

Summary of cost-effectiveness

Six studies were identified that met our criteria. However, none were from a British society or UK NHS perspective. The evaluations showed that the cost-effectiveness of CRT increases as the time horizon lengthens.

Industry submissions to NICE contained four cost-effectiveness analyses, of which two were more appropriate as reference cases for this report. One, based mainly on the European CARE-HF trial but using resource use data from a UK hospital, used a discrete event simulation model. This gave estimated incremental cost-effectiveness ratios (ICERs) of £15,645 per QALY for CRT-P versus OPT. The other analysis was based on the results of the COMPANION trial, but substituted UK-based hospital unit costs for those originally based on US-defined diagnostic related groups. They estimated an ICER of £2818 per QALY gained by CRT-P versus OPT and a cost per QALY gained of £22,384 for CRT-D versus OPT. Neither of these analyses had time horizons that would adequately include the periodic cost of unit or device replacements (e.g. due to battery depletion), and neither directly compared the cost-effectiveness of CRT-D with CRT-P.

Compared with OPT, the PenTAG Markov model base case analysis (over a lifetime) estimated that CRT-P conferred an additional 0.70 QALYs for an additional £11,630 per person, giving an estimated ICER of £16,735 per QALY gained for a mixed age cohort (range £14,630–20,333). CRT-D versus CRT-P conferred an additional 0.29 QALYs for an additional £11,689 per person, giving an ICER of £40,160 per QALY for a mixed age cohort (range £26,645–59,391).

The PenTAG ICERs are higher than those from the industry-submitted analyses. These differences are due to the industry analyses having higher estimated QALYs and failing to include the costs incurred from repeated replacement of devices, due to modelling the decision over 5 years.

Sensitivity analysis

One-way sensitivity analyses showed that the PenTAG model was sensitive to model time horizon, device lifetime, discount rate applied to health benefit, probability of a major arrhythmic event, risk of sudden cardiac death and the risk of death from worsening heart failure.

Probabilistic sensitivity analysis based on 1000 simulated trials showed that, at a willingness-to-pay (WTP) threshold of £30,000 per QALY:

• CRT-P versus OPT: CRT-P was likely to be cost-effective in 91.3% of simulations and CRT-P was negatively dominated (i.e. the more you pay, the less quality of life you receive) in 0.4% of simulations.
• CRT-P versus CRT-D: CRT-D was likely to be cost-effective in 26.3% of simulations and CRT-P dominated CRT-D in 7.8% of simulations.

The relative risk for risk of sudden cardiac death when CRT-D is compared with OPT is 0.44 in the base case. This treatment becomes cost-ineffective, at a WTP threshold of £30,000 when this value is greater than 0.65.

When both CRT-P and CRT-D were considered as competing technologies with each other and OPT (three-way probabilistic analysis), and at the same WTP threshold, there was a 68% probability that CRT-P provided the highest expected net benefit. The WTP threshold would need to be above £40,000 before CRT-D provided the highest expected net benefit.

Strengths, limitations of the analyses and uncertainties

The strengths of this systematic review and economic evaluation are that it is comprehensive, systematic, up to date and independent.

The limitations of the clinical systematic review were that evidence directly comparing CRT-P with CRT-D was limited, follow-up times were short (three trials 6 months, one trial 15 months, one trial 36 months) and the studies were not sufficiently powered for subgroup analyses. People with atrial fibrillation or an indication for an ICD were excluded from the studies.

The main limitation of the review of economic evaluations, for our purposes, was that none of the studies were from a UK NHS or British society perspective.

The submission from industry was based on European and US trials and case series data from a UK hospital. CRT-P was not compared with CRT-D.

Generalisability of the findings

The populations in the included trials may not be fully representative of the general population of people with heart failure in UK; typically, the trial populations were younger and had less co-morbidity. In addition, none of the trials included in this systematic review included people with conventional indications for an ICD.

Implications for service provision

The rate of implanting CRT devices in UK is currently increasing by about 50% per annum. There are consequent implications for the training of cardiologists and related clinical staff and the adequate provision of implantation centres and associated diagnostic infrastructure.

Recommendations for future research

The following areas are suggested for further research.

• Prediction of non-responders: systematic reviews of current evidence and further primary studies are needed in this area.
• Appropriate use of CRT-D devices: only the COMPANION study directly compared CRT-P and CRT-D devices. The question remains as to which group of heart failure people should receive a CRT-D device.
• NYHA classes I and II: RCTs powered to detect differences in mortality and heart failure hospitalisation are needed.
• Long-term safety data: observational studies are needed to determine the long-term safety of CRT devices.
• Based on the expected value of perfect parameter information analysis, studies with long-term follow-up are needed to provide a better understanding of the different modes of death in people with a CRT device and also those receiving OPT. The results indicate that information about all hazard ratios would give the maximum reduction in decision uncertainty.