Discussion

Statement of principal contributions

Cardiovascular disease microsimulation policy model

A new UK CVD microsimulation policy model has been developed to predict an individual’s lifetime risks of CVD and death. In addition to sociodemographic and clinical characteristics, the model incorporates socioeconomic status and modifiable lifestyle factors (physical activity, diet quality and BMI) and tracks medical history and comorbidities. It has shown good accuracy across various categories of the UK population. Moreover, the model includes individualised predictions of HRQoL associated with individuals’ characteristics and adverse events. A significant decrease in QoL during the year of acute events, with partial recovery of QoL in subsequent years, except for stroke and longer diabetes duration, is reported. The study also quantified the temporal impacts of cardiovascular events on UK primary and hospital care costs, revealing substantial excess primary and hospital care costs during the years with events compared with the years prior to event. However, although the excess hospital care costs decreased substantially in the years following the event, the excess primary care costs remained relatively stable. The model’s web interface includes all the features mentioned, providing lifetime individual-level projections that can help inform economic assessments of the value of health interventions to reduce CVD risk.73 The model could be particularly useful for a cost-effectiveness assessment of CVD prevention treatments such as lipid- or BP-management therapies.

Statin cost-effectiveness findings

We conducted an assessment of the cost-effectiveness of statin therapy for men and women aged ≥ 40 years in the UK, taking into account the contemporary CVD incidence and mortality and the potential trade-offs between statin benefits and risks. We considered sex, age, previous CVD status or estimated CVD risk, and pre-treatment LDL-C level, the key patient characteristics by which to present our findings. Our findings show that lifetime statin therapy increased QoL-adjusted survival and was highly cost-effective in all patient categories studied. Higher-intensity statin therapy was also cost-effective in many categories of patients at higher CVD risk or higher LDL-C levels. These results remained robust in sensitivity analyses and at lower cost-per-QALY-gained cost-effectiveness thresholds.

There is greater uncertainty about the value of statin treatment for people ≥ 70 years old. A recent IPD meta-analysis noted trends across age categories towards smaller proportional reductions in major coronary events and VDs per 1.0-mmol/l reduction in LDL-C with statin therapy.5 Data were particularly limited among participants > 75 years old without previous CVD, for whom there was no direct evidence for statistically significant CVD risk reductions with statin therapy. In this study we included three scenarios for size of statin effects in the cost-effectiveness assessment among older people. First, we used the overall relative risk reductions in cardiovascular events per 1-mmol/l LDL-C reduction with statin therapy. Second, we used the relative risk reductions in cardiovascular events only among participants aged > 75 years at time of randomisation in trials.5 Finally, we used even smaller relative reduction in cardiovascular events per 1-mmol/l LDL-C reduction with statin therapy among older people without previous CVD.5 Despite smaller net health benefits in the later scenarios, statin therapy remained cost-effective, with standard statin therapy preferred in the latest scenario, although the level of uncertainty was substantially higher. Our findings contradict the observation in an earlier study of cost-effectiveness of statin therapy for the primary prevention of CVD in people ≥ 75 years old,74 in which the authors reported that even a small hypothetical increase in a geriatric-specific adverse effect (i.e. reducing QoL by 0.003–0.004) would offset the cardiovascular benefit of statin therapy. In this study, we explicitly integrated the known small excesses of myopathy, rhabdomyolysis and increased treatment intensity-dependent hazard of incident diabetes with statin treatment. In addition, the results remained robust to further hypothetical statin-associated reductions in QoL up to 0.005 per year, as well as to lower effectiveness of statin treatment in older people, suggesting that the uncertainty in the value of statin therapy for older people may be more limited than previously suggested.

Our research suggests that the entire UK population aged ≥ 40 years would benefit from statin therapy cost-effectively, including the 56% of those aged 40–70 years (or 14 million individuals), many of whom have raised LDL-C levels, who are not currently recommended for statin therapy.75 Patients at only moderately raised CVD risk who have higher LDL-C levels are likely to derive benefits comparable to or exceeding those in patient categories currently treated. Our results could be particularly useful in discussions between patients and doctors as part of shared decision-making. More efforts are needed to ensure clear communication of the evidence to doctors, individual patients and populations groups, and the efficient integration of statin treatment into the health care of a wider range of population categories. While initiating statin treatment is cost-effective also in the presence of treatment discontinuation, long-term persistence with statin therapy is crucial to maximise benefits.

Our findings present a dilemma for policy-makers faced with constrained resources. For example, it is uncertain whether UK primary care has the resources to expand services at the same level of per-person cost of initiating and monitoring statin treatment. Nonetheless, our findings are robust to some increases in costs. For example, at current costs, the incremental cost per QALY gained with statin treatment is much lower than the threshold of £20,000–30,000 per QALY gained used for new treatments.64

Strengths of the assessment

The key strengths of the study include the use of high-quality evidence and robust methodology. The CVD microsimulation policy model was derived using large and rich IPD with substantial duration of follow-up, enabling us to reliably model the disease risks, QoL-adjusted survival and healthcare costs of individuals over time using their characteristics at entry. The model integrated the underlying risk of incident diabetes, which enabled us to project the excess diabetes risk with statin therapy and its consequences, as well as the risk of incident cancer, the largest competing risk affecting HRQoL and survival of individuals, in parallel with any benefits from CVD risk reductions. The ability to report results for distinct categories of actual person profiles and fully assess parameter uncertainty is a further strength.

This cost-effectiveness study of statin therapy has key further advantages over previous studies. Effects of statin therapies were informed by an individual participant meta-analysis of all large statin trials. It improves on the CVD decision-analytic modelling informing the current NICE guideline of statin treatment to reduce CVD risk15 by using recent IPD and integrating all known adverse effects of statin therapy. Recently, Kohli-Lynch et al.76 reported population-level cost-effectiveness results in Scotland, using recent population data, at different thresholds of CVD risk, CVD risk and LDL level, and age. The present study instead reports detailed results in distinct UK population categories and for both standard and higher-intensity statin regimens, directly informing decision-makers’ trade-offs at different cost-effectiveness thresholds and presenting compelling evidence that, at currently stated thresholds, initiating statin therapy could be considered good value for money for all UK adults ≥ 40 years old. This study improves on the 10-year results reported by Greving et al.77 by extending the time horizon to a person’s lifetime and reaches different conclusions about the cost-effectiveness of statin therapy. In general, the results over a limited time horizon, despite avoiding the assumptions needed for extrapolation, result in a downwards bias of the estimated benefits and cost-effectiveness of preventive treatments that reduce disease risks and should be discouraged. Recent US cost-effectiveness analyses of statin therapy, although based on older data, also reported evidence for broader statin cost-effectiveness in adults, largely irrespective of level of CVD risk, but sensitive to level of possible ‘disutility’ of taking a daily statin pill.61,78,79 Our analysis suggests that this sensitivity is likely to be relevant only among people at lower CVD risk or with lower LDL-C levels.

Limitations of the assessment

The study has several limitations.

First, while we explicitly modelled the effects of evolving cancer and diabetes risks, further morbidities likely to accumulate in older age, together with the need for co-medication, were beyond the scope of the current project. While the increasing hazard of NVD captures some of these effects, and the scenario analyses (e.g. higher cancer risk with statin therapy, further QoL disutilities) provide assurance that these are unlikely to materially affect statins’ value, further work to assess the implications of multimorbidity and polypharmacy will be helpful.

Second, the risk equations used in our framework were derived and validated in population cohorts, such as UKB and Whitehall II, which may limit the generalisability of our findings due to the potential presence of a healthier volunteer effect. The under-representation of ethnic minority groups in these cohorts is a further limitation. However, the individual equations include a broad range of characteristics to account for the effects of lifestyle, socioeconomic and clinical factors, enabling us to extend findings to populations with varying distributions of these characteristics. Nevertheless, some selection bias may persist, and scenario analyses are employed to further check for the impact of this on study findings.

Third, in the analytical framework we modelled the first occurrences only of the disease events of interest. While subsequent disease events from same type were not explicitly modelled, the use of the same population data to model risks of further events, in particular vascular and NVD risks, would have also accounted for the impacts of subsequent events (as both risk factors and first occurrences of events contribute to subsequent risks of events as well as deaths).

Fourth, a significant limitation of our cost-effectiveness results is the currently limited direct evidence for the effects of statins in older people without previous CVD and the absence of randomised data on the long-term effects of statin therapy. Nonetheless, we found that our results were not highly sensitive to hypothesised lower efficacy in sensitivity analyses. While ongoing large, randomised trials in older primary prevention populations will improve the evidence, our findings’ general robustness to variations in key parameters suggests that delaying statin treatment for millions of older people while awaiting new evidence is not justified.

Fifth, our analytical framework was developed and assessments were performed for individuals aged ≥ 40 years, so we are unable to provide information on the effects of initiating statin treatment on younger individuals.

Finally, developing a new CVD model using contemporary population data was motivated by the need to reflect current disease risks in view of decreasing cardiovascular mortality and incidence over recent decades. While the findings in this report are likely to describe well the current disease risks, further changes in disease risks will necessitate an update of the model framework. The approach taken in the present study could be helpful in calibrating the reported risk equations with new population data.

Equality, diversity and inclusion

Participant representation

Reflect on the important population characteristics for your study. What active steps were taken to optimise participation of relevant people?

This study aimed to provide an effectiveness and cost-effectiveness assessment of statin treatment in people ≥ 40 years old in the general UK population. With this focus, we used the data from participants (≥ 100,000) in the trials included in the CTTC to inform the treatments’ effect as well as an initial estimation of disease risks and disease progression, and we used the data from participants (aged 40–70, ≥ 500,000) in the UKB to further inform the disease progression and healthcare costs, and used data from the HSE (nationally representative) to inform the QoL assessment. As we do not exclude specific populations from the analysis, all types of people were included in our research. As the UKB participants could be healthier than the general UK population, we also checked our projections using follow-up data on participants in the Whitehall II study, a cohort of British civil servants. We present results across categories by age, sex, CVD history or CVD risk and LDL-C level that are more relevant to the corresponding categories in the general UK population.

Was the participant population inclusive? If not, an explanation should be provided as to what prevented this or any improvements that could be made for future study designs

We believe that the participant population is reasonably inclusive and can answer the question relevant to general UK population. However, we relied on available IPD from UK cohorts. We acknowledge an under-representation of minority ethnic groups in UKB. Future studies could consider including further population data to better represent the UK general population, such as the UK Clinical Practice Research Datalink databases, despite there being more limited data on participant characteristics in these data sets.

If applicable, provide comments on how gaps in prior evidence/data were identified and how were these addressed during the project. Did this research uncover any gaps in knowledge/data during the project; how could this be addressed in future research?

As noted in the 2014 NICE guideline ‘Lipid modification: cardiovascular risk assessment and the modification of blood lipids for the primary and secondary prevention of cardiovascular disease’,80 the prior economic evidence is limited in its methodological ability to predict the impact of statin treatment on an individual’s disease risks over a lifetime. This prior evidence is also limited in its ability to project QoL of and costs for individuals with different characteristics. We addressed these issues by using the IPD from large statin trials to allow a reliable estimation of disease progression using adjudicated outcomes in the trials; using a large population cohort to calibrate the disease models and inform the healthcare costs; and using national population surveys to inform QoL assessment. Future research should focus on estimating an even more detailed set of outcomes and using a broader population with longer follow-up data to support the modelling.

How have you made sure text and pictures, used as part of participant and public engagement, were inclusive and accessible and not discouraging to under-represented groups?

Our public representatives’ members on the project management and oversight groups helped us identify the key concern of the general public and helped us develop project outreach materials that were clear and conveyed well the key message from the project to the general public. In particular, we presented results across diverse patient categories and focused on clarity in our presentation, which involved simplifying text and figures.

Reflections on your research team and wider involvement

Does your research team include those from groups who are generally under-represented in your field of study? If not, please state why this is the case.

The research team was reasonably broad and included good representation of areas of expertise, gender and national identities. Unfortunately, the team did not include representation of key minority ethnic groups directly and instead we relied on discussing our findings more broadly, including with the Oxford Population Health (OxPop) Public Advisory Panel to improve our coverage of broader and more diverse views and opinions.

Was there a range of experience and expertise across the research team? How were development opportunities provided, particularly for more junior members of the team?

Our research team included people with ranges of experiences and expertise, from a postdoctoral researcher to senior postdoctoral researchers, associate professors and professors. They are from different backgrounds, including health economists, epidemiologists, statisticians, a pharmacologist, clinicians, and patient and public representatives. More junior researchers were well supervised, benefited from advice from more experienced project members, and had a range of training opportunities in specialist and general research skills and development.

Were participants and members of the public involved representative of the disease and geographic areas? What role/s did they provide and how did they benefit the project?

Three patient and public representatives were included in the study, and we also presented our work to a broader public panel consisting of 13 individuals who represent the target population and geographic areas affected by the disease. Their input was invaluable in improving the dissemination of the study results and helping them understand the scientific work behind the messages sent to the public. They were also the first group to learn about the results from our project, which provided them with the opportunity to benefit from the project.

Patient and public involvement and engagement

We reported the patient and participant involvement following the structure of the GRIPP2 Short Form (see Report Supplementary Material 1).81

Aim

We aimed to include patient and public involvement representation throughout the project to make the project more relevant to the patient group and the wider public and to facilitate the dissemination of project results.

Methods

We involved two members of the public from the start of this project. The first member (PPI-A1) was expected to be involved in all nine meetings of the project management group to discuss the progress of the project every 3 or 4 months until the end of the project. The second member (PPI-B) was expected to be involved in all the three meetings of the project oversight group to discuss the project’s progress and emerging results yearly until the end of the project. In the last year of the project, we included a further PPI member, PPI-A2, to replace PPI-A1 in the management group, as PPI-A1 stepped down due to ill health. The PPI members had some direct experience with the statin intervention as well as some indirect experience with CVD in relatives.

During the project, we held further separate meetings with our PPI members to discuss the study methodology, model outline and, later, the emerging results and solicit their feedback on whether (1) we included all important patient characteristics and outcomes; (2) we used language that is easily understood by the patients and the public; and (3) the presentation of the emerging results was relevant to the patients and the public. With their feedback, we modified the model and further developed the dissemination slides of the results of the cost-effectiveness study presented to the OxPop Public Advisory Panel, which included 13 patient and public representatives with diverse ages, ethnicities and backgrounds, for further feedback. At these meetings the final versions of the slides were formed for further dissemination of the results to the general public.

Results

During the project management group meetings, PPI-A1 expressed concerns about the muscle side effect of statin. During the Project Oversight Group meetings and in separate PPI meetings, PPI-B noted that the models omitted some important individual factors, including lifestyle (e.g. physical activity and diet quality) and mental health status. These outcomes and factors were subsequently captured in the model development and in the cost-effectiveness analysis of statin therapy. PPI members also appreciated model validation and that the model was able to predict rates of disease events corresponding to observed rates in external data.

The separate meetings with our PPI representatives helped us to shape the slides for the presentation to the OxPop Public Advisory Panel. We first presented our emerging results from a scientific researcher’s perspective to PPI-B. With feedback from PPI-B and PPI-A2, we reduced the number of slides from 56 to 13, restricting the content to what would be relevant to and understandable by the patient group and the general public. We modified our research messages into plain language, such as mentioning that the LDL-C reduced by statin was the ‘bad’ cholesterol, while HDL-C was the ‘good’ cholesterol; referring to level of health with respect to QALYs (e.g. 1 QALY = 1 year in good health); and using value-for-money when referring to cost-effectiveness.

Patient and public involvement member A2 helped to make sure that we addressed all key questions and considered how to enhance the project dissemination, including (1) what did we say we would do in the protocol for this project?; (2) can we learn or use the ideas from similar projects?; (3) what are the key messages from the project?; (4) who are the target audiences?; and (5) who are our partners and allies?

With these questions in mind, we refined the slides for the presentation to the OxPop Public Advisory Panel to showcase the health benefits and value-for-money of statins in the present-day UK. The slides focused on the main results that are expected to be of interest to a general audience.

The panel understood well the results indicating the benefits of statins and provided further helpful comments to improve the presentation of results to the general public, such as a suggestion to present statins’ benefits in different time frames as well as lifetime.

Discussion and reflections/critical perspective

Our PPI members provided important contributions to the study design and made an invaluable contribution to disseminating the research results to the general public using language that can be easily understood and to restrict the results to only those relevant to the general public.

Dissemination of the results of the study

We aimed to disseminate our findings to audiences with different backgrounds, including academia, research community, trainees, policy-makers, people from the industry, patients and the general public.

Table 21 summarises the undertaken project dissemination activities.