Introduction

Background and definition of the clinical problem

Statins are a class of lipid-lowering medications that have been reliably shown in large randomised controlled trials to reduce the risk of heart attacks, strokes and vascular mortality in different categories of patients. Since the 1990s, the Cholesterol Treatment Trialists’ Collaboration (CTTC) has co-ordinated a prospective meta-analysis of all large statin trials.1 This work has demonstrated that statins produce similar proportional reductions of about one-quarter in cardiovascular disease (CVD) risk per 1-mmol/l reduction in low-density lipoprotein cholesterol (LDL-C) in a wide range of people [e.g. men and women; older and younger people; people with and without previous CVD; people with high and low CVD risk; people with diabetes or non-dialysis-dependent chronic kidney disease (CKD)], and that further reductions in LDL-C with more intensive statin therapy produce further reductions in CVD risk.2–8

There is some controversy, however, over the size of statin therapy’s net benefits for people at low CVD risk, and for elderly people, and there are concerns about the safety of statins. Many of the safety concerns originate from non-randomised studies9,10 and may be unreliable.11 Nevertheless, in randomised trials, standard statin dose regimens have been associated with a proportional increase of about 10% in incident diabetes,12 and more intensive statin regimens with about twice this risk.13 Such adverse effects should be considered in evaluating the net effects of statins particularly in people at low CVD risk.

Cost-effectiveness analyses help decision-makers obtain better value for money by targeting healthcare resources at interventions and population groups where the net health gain is greatest in relation to the net cost. However, questions remain as to how such analyses should be performed, which leads decision-makers to question their reliability. Decision-analytic models, typically developed using summary data from multiple sources, and (summary) treatment effects from randomised clinical trials, are commonly used but are likely to be of limited reliability in particular categories of people defined by multiple characteristics (e.g. by disease risk, age, sex, comorbidities).

Rationale for the study

Statins are now cheap14 and the direct costs to the NHS are less of a concern, but the efforts required to initiate and support people on treatment should not be understated. In England and Wales, the National Institute for Health and Care Excellence (NICE) recommends treatment with high-intensity statin therapy for all individuals with a history of previous CVD or diabetes and, since 2014, treatment with medium-intensity statins for those without such a history who have a 10-year CVD risk ≥ 10%.15,16 Furthermore, this guidance also states that statin treatment should not be ruled out if the estimated 10-year CVD risk was ˂ 10% if the patient had an informed preference for taking statin or their risk might be underestimated, opening the possibility for wider statin use but stopping short of concrete guidance. The economic evaluations of statin therapy for CVD prevention in these guidelines followed the conventional approach of decision- analytic modelling using published summary data. An evaluative cost-effectiveness framework based on summary data, however, does not allow for reliable assessment of disease risks and treatment effects over time or in categories of patients. Furthermore, in the absence of an assessment of model validity, the reliability of the results is unclear.

The NICE guidance affects a large section of the population (about 37% of 30- to 84-year-olds17). Unreliable cost-effectiveness results could lead to the consequence that people who do not derive worthwhile health benefit from the statin treatment are recommended for the treatment, while people who could derive worthwhile benefit from the statin treatment are not recommended for the treatment or are recommended for a suboptimal intensity of statin treatment. The uncertainty in evidence can also affect the level of implementation of guideline recommendations with ample evidence indicating suboptimal statin use among people recommended for treatment18–20 with both individual patient and prescriber factors likely contributing.

Consequently, in 2014 a research recommendation was made in NICE CG181 for the development of a cost-effectiveness analysis of statins informed by the individual participant data (IPD) of randomised clinical trials.16 This recommendation was taken forward in this project. The following priority research areas were identified: (1) use of IPD from randomised clinical trials of statins to develop more detailed and reliable (e.g. based on time-to-event analysis) cost-effectiveness analyses and (2) use of such analyses to produce detailed results for the effectiveness and cost-effectiveness of statin therapies in categories of patients by CVD risk and other patient characteristics (e.g. age, sex, comorbidities). These analyses are timely in view of the increasing availability of statins (i.e. all widely used statins are now available generically in the UK); the growing evidence for the effectiveness of new treatments [e.g. ezetimibe, PCSK9 (proprotein convertase subtilisin/kexin type 9) inhibitors]; and the developments in UK CVD risk scoring (e.g. QRISK^®3 10-year CVD risk score21).

Cost-effectiveness studies embedded within individual statin trials have demonstrated an ability to evaluate the cost-effectiveness of statin regimens in categories of patients.22–26 By combining IPD-based multivariable time-to-event disease risk equations with estimates of the relative effects of statins on disease risks into interlinked disease models, such frameworks have evaluated statins’ cost-effectiveness reliably in particular categories of participants in the trials (e.g. by CVD risk, age and gender).23,25 The CTTC database is a unique resource for the development of further, more detailed analyses to guide statin recommendations for individual patients. It includes IPD from large statin trials among nearly 180,000 participants with well-documented baseline characteristics and first occurrence of major vascular, cancer and mortality events during the studies’ follow-up periods. However, trials in the database recruited their participants from different countries during the period 1980s to 2010 and, therefore, participants’ disease risks may potentially not generalise well to the contemporary UK population.

In this project we aimed to strengthen the evidence about the cost-effectiveness of statins using the CTTC IPD together with large contemporary UK population cohorts. The cost-effectiveness analysis of statin regimens developed using these data would account for the timing of disease events and the beneficial and adverse effects of statin therapies and allow for more reliable estimates of the net effects of statin therapies of different intensity on quality-adjusted life expectancy and healthcare costs in categories of patients.

The research question addressed in this project was ‘What is the cost-effectiveness of different statin therapies in different categories of people?’. We assessed the effects of standard statin therapy, defined as a therapy that achieves about 35–45% LDL-C reduction (e.g. atorvastatin 20 mg/day, rosuvastatin 5–10 mg/day or simvastatin 40–80 mg/day), and of higher-intensity statin therapy achieving ≥ 45% reduction (e.g. atorvastatin 40–80 mg/day, rosuvastatin 20–40 mg/day) (Table 1).

Changes to methods

The project was developed following a prospectively developed study protocol. Protocol versions 1 (April 2019) and 2 (November 2022) are available at https://fundingawards.nihr.ac.uk/award/17/140/02.28 Three changes were made from version 1 to version 2 of the study protocol. First, the decision not to model recurrent CVD events of the same type was taken because such events were not available in the CTTC database29 at the onset of the project and the subsequently developed microsimulation model was judged to be of sufficient depth and to have an excellent performance in key participant categories. Second, we did not use the Heart Protection Study long-term follow-up data30 in the microsimulation model validation as the model validation in the UK Biobank (UKB) cohort31 indicated a need to calibrate the model in a contemporary population cohort such as UKB. Third, the decision was taken to use the UKB resurvey data to derive a cohort of people aged ≥ 70 years to guide the development of the assessment of value of statin therapy in older people.

Subsequent to protocol version 2, there have been only limited changes. First, the Whitehall II study data contributed more substantively to the development of the statin cost-effectiveness study in older people. In particular, we used the Whitehall II study32 data from Phase 9 onwards (2007–9 survey with 10–12 years’ follow-up thereafter) to validate the performance of the microsimulation model both in people 40–70 years and in people ≥ 70 years old. Second, to increase the size of the older people cohort, we used the UKB and Whitehall II studies’ data for participants ≥ 70 years old together, instead of UKB data alone, to project and present results for cost-effectiveness of statin therapy in older people. Finally, we decided against reporting results for cost-effectiveness of statin therapy by lifetime CVD risk (QRISKlifetime33), as statin therapy was already shown to be highly effective and cost-effective across all categories of people aged ≥ 40 years.

Organisation of the report

This report is compiled using a number of chapters presenting different components of the project.

Chapter 2 outlines the initial model development using the IPD of large statin trials in the CTTC database to estimate key CVD and nonvascular disease risk equations.

Chapter 3 describes the CVD microsimulation policy model calibration, further development, and validation in the UKB database and in the Whitehall II study.

Chapter 4 presents the hospital and primary care cost regression models, estimated using the UKB cohort and linked hospital admissions and primary care data, allowing an assessment of the healthcare costs associated with individual patient characteristics and history of vascular and nonvascular disease events.

Chapter 5 reports the quality-of-life (QoL) regression model, developed using IPD from the Health Survey for England (HSE) in 2006, 2011 and 2017, allowing an assessment of health-related quality of life (HRQoL) associated with individual patient characteristics and history of vascular and nonvascular disease events.

Chapter 6 reports our findings for the cost-effectiveness of statin therapy in categories of people 40–70 years old using the CVD microsimulation policy model.

Chapter 7 reports further validation of the CVD policy model among people aged ≥ 70 years in the UKB and Whitehall II cohorts and our findings for the cost-effectiveness of statin therapy among people aged ≥ 70 years in the UK.

Chapter 8 describes the web-based interface to the CVD microsimulation policy model to facilitate external use of the model.

Chapter 9 presents a general discussion and Chapter 10 summarises the key findings of this project.