Improving evidence for economic evaluation and healthcare policy in relation to long term conditions using Mendelian Randomization on large cohorts

Healthcare funders have limited budgets. Choices must be made between spending money on different health conditions, and on the different treatments available for each health condition. An important consideration in these decisions is the cost-effectiveness of interventions. Cost-effectiveness reflects the additional value that can be obtained from a treatment relative to a comparator treatment, taking into account the costs of each treatment. Valid evidence of this kind can be difficult to obtain. My fellowship will apply new techniques recently developed in the study of genetics to generate robust evidence on two long-term conditions that can be used in assessing treatment cost-effectiveness.

Much available research cannot meaningfully identify the long-term consequences of health conditions for healthcare cost and quality of life. This is because observational studies, which describe associations between health conditions and outcomes, are prone to bias. For example, patients with one condition may have other conditions, and it is difficult to establish which condition has the greatest influence on quality of life. It is difficult to distinguish the effect of conditions from other factors that affect quality of life, such as employment status. Changes in quality of life may be a cause or consequence of health conditions. Measurement of health is complicated if symptoms are unreported or measured with error. Randomized controlled trials, based on fair comparisons between similar individuals, can overcome these problems, but long-term trials may not be affordable, feasible or ethical.

I propose to use information on the relationship between genetic variants and health conditions to overcome these problems. Genetic variants refer to pieces of the genetic code that differ among individuals in a population. These variants occur in a near random manner at the time of conception. Some variants are known to influence health conditions, such as obesity or heart disease. In these cases, the variants are useful proxies for the condition itself, since they are not affected by the confounding factors that afflict observational studies, may be unrelated to other conditions, and are precisely measured.

Variants meeting these desirable characteristics can be used in statistical models in a manner that approximates a randomized controlled trial. This allows meaningful comparisons of the effects of health conditions on healthcare cost and quality of life to be made. These comparisons can be used in evaluations of treatments and other interventions that may have long-term effects. This evidence is also relevant to health policy more generally, since the management of long-term conditions poses challenges because of their prevalence, impact on patient health and the wider economy.

Data to undertake these studies will be taken from two kinds of data source. The first will be very large studies - genome wide association studies - that measure the relationship between genetic variants and specific health conditions. For example, almost 100 genetic variants are now known to influence the risk of obesity. The second will be large datasets that link genetic, health condition and medical record data (e.g. 500,000-person UK Biobank) and/or quality of life data (e.g. the English Longitudinal Study of Ageing).

I will study the effects of two complex health conditions- obesity and coronary artery disease - on healthcare cost and on quality of life using UK and US data. These conditions could be considered exemplar long term conditions: they are believed to be associated with large costs, have big impacts on patients and are prevalent.

My fellowship will be the first such application of these methods to fundamental questions concerning cost and quality of life. The research will have important implications for the evaluation of new and existing health technologies, for healthcare policy, and the setting of research priorities

Background: Accurate measurement of the effects of health conditions on healthcare cost and quality of life is important in the pragmatic evaluation of interventions and for healthcare policy more generally. However, most existing evidence is based either on short-term effect estimates from RCTs, or observational methods prone to bias. The proposed work will produce new and robust evidence of the causal effect of prevalent health conditions on cost and quality of life.
Objectives: To apply the techniques of Mendelian Randomization in order to produce robust causal estimates of the effects of two exemplar, prevalent and complex conditions (obesity and coronary artery disease) on healthcare cost and on quality of life. To improve population health by generating more robust evidence for the evaluation of interventions, and to increase awareness of the proposed methods.
Methods: The causal effects of obesity and coronary artery disease on healthcare cost and quality of life will be estimated by using data on genetic variants (single nucleotide polymorphisms) as proxy or instrumental variables, since appropriately selected genetic variants will be approximately randomly allocated to individuals (no confounding), at the time of conception (no reverse causality), and with can be measured with high precision (little measurement error).
Data: I will use data from well-conducted genome-wide association studies to obtain robust evidence of the associations of genetic variants and health conditions (e.g. the CARDIOGRAM consortium for coronary artery disease). I will obtain data on the association of variants and costs from the UK Biobank. I will obtain information on the association between variants and quality of life from the Wisconsin Longitudinal Study, the Health and Retirement Study, and the English Longitudinal Study of ageing.
Outputs: Research papers, methodological expositions and other dissemination outputs including conference presentations and public engagement

Who will benefit from this research?
The public: The ultimate aim of the research is to improve population health by increasing the quality of evidence used by decision makers. Moreover, a clearer appreciation of the consequences of health conditions for costs and quality of life is likely to considerable interest to patients and their families and carers.

Policy makers: Improved evidence from cost-effectiveness analysis is fundamental to the activities of NICE and other similar international bodies, and should help increase the effectiveness of allocation decisions. I expect the results to be relevant to the Department of Health, possibly the Treasury, and similar organisations elsewhere. For example, the influential 'Foresight' report on obesity, produced under the direction of the Chief Scientific Adviser to the UK government, described the costs of obesity using data drawn from primarily observational analyses. Better causal evidence will improve the robustness of future policy documents.

Healthcare funders commissioners: The work is of fundamental relevance to the activities of funders and commissioners. For example, the evidence I will produce could help an insurer estimate by how much insurance premiums should change in response to a reduction in the incidence of obesity and therefore in obesity-related healthcare cost. Commissioners could, for example, use the evidence on the healthcare costs causally associated with coronary artery disease to evaluate the cost-effectiveness of screening programmes or other interventions.

Private sector: There is evidence (e.g. Nelson et al, Nat Gen, 2015) that drugs and indications can be better targeted when account is taken of genetic influence. Mendelian Randomization analysis in epidemiology has already identified or dismissed possible therapeutic targets, for example C-reactive protein in the case of long-term cardiovascular disease. In addition to supporting higher-quality private sector submissions to NICE, the proposed methodology could support targeting of intervention development, since it could identify where therapeutic development is likely/unlikely to be both clinically effective and cost-effective. Improved causal evidence will also be relevant to health insurance companies.

Public or third sector: Improved causal evidence will clarify resource needs for different disease areas, and will support targeted priority setting.

HOW WILL THESE PARTIES BENEFIT FROM THIS RESEARCH?
The fundamental motivation of the work is to improve the evidence used to inform the allocation of scarce healthcare resources in order to improve population health in the UK and elsewhere. The research will introduce a new methodology for health economists that can be used to support decision making, the evaluation of interventions and policies, and the identification of targets for therapy development. The timing of impacts will be a function of the rate at which these methodologies are applied by researchers to different disease areas, and how quickly the methodologies are adopted in guidelines and policy decisions. I describe elsewhere (in relation to Impact and Communication) the activities that I will undertake to support rapid dissemination.

The use of valid causal evidence has the potential to optimise medical expenditures, broaden the scope of therapies available to patients, and ultimately improve therapeutic value. This will support economic competitiveness via a beneficial impact on healthcare innovation, the promotion of which is one of the objectives of the UK Government's 'Strategy for Life Sciences'

I aim to consolidate and develop existing and new skills, particularly in topics at the intersection of genetic epidemiology and health economics. I intend to undergo the rigorous training programme described elsewhere, to develop national and international collaborations, to secure impact and to increase my independence as a researcher