Methods for predicting vascular risk: a rapid review of the evidence
Author: Dr Mary Webb, Public Health Practitioner
Date: 27 October 2010 / Version: 1
Publication/ Distribution: Public Health Wales Vascular Review Group
Review Date: N/A
Purpose and Summary of Document: A systematic coordinated risk assessment programme for coronary vascular disease (CVD) is a high priority for NHS Wales. The Health Check Programme in England recommends universal screening for CVD risk for all people aged 40 – 74 years and is costly. Pre-selection of patients using data already collected in general practice has been suggested as a cost effective alternative to the Health Check strategy of universal screening for those aged 40 – 74 years. There are many risk prediction tools in existence and the choice of method for risk assessment has been subject to intense investigation. Public Health Wales were asked by the Welsh Assembly Government to review the evidence on tools and screening methods for vascular risk assessment.
The most frequently used tool is the Framingham model. This has been shown to have disadvantages and other tools have been developed. The review of the evidence, which was in the main observational or expert opinion, suggests that the QRISK2 tool has advantages over the Framingham tool (NICE no longer recommends the Framingham equation). There is preliminary evidence from cohort and modelling studies which indicates that this may be an effective method for vascular risk assessment. Cost-effective data is lacking.
Work Plan reference: HS38
Contents
Executive summary / 3
1. Introduction / 6
2. Aims / 7
3. Research questions / 7
4. Methods / 8
4.1 Identifying existing and ongoing research / 8
5. Results / 9
5.1 Question 1 / 9
5.2 Question 2 / 10
5.3 Question 3 / 12
5.3.1 Risk tools/methods / 13
Framingham model / 13
Modified Framingham models / 15
Alternatives to Framingham / 15
SCORE (Systematic Coronary risk evaluation) / 15
ASSIGN (Assessing cardiovascular risk using SIGN guidelines) / 16
QRISK / 16
5.3.2 Incorporation of new risk factors into risk estimation systems / 23
5.3.3 Guideline recommendations for risk assessment methods / 27
5.3.4 Pre-stratification of patients using data collected routinely in primary care / 31
5.4 Question 4 / 37
6. Conclusions / 39
7. References / 40
Appendix 1 Evidence levels and quality grading (modified from NICE Guideline Methodology Manual) / 45
Appendix 2 Evidence table / 46
© 2010 Public Health Wales NHS Trust.
Material contained in this document may be reproduced without prior permission provided it is done so accurately and is not used in a misleading context.
Acknowledgement to Public Health Wales NHS Trust to be stated.
Executive summary
A rapid review of the evidence on tools/methods for vascular risk assessment was performed using previously described and validated methods. The term vascular risk was used to denote the risk of coronary heart disease, diabetes, stroke and renal disease.
The review aimed to find evidence for the priorities listed in the report from the Welsh Assembly Government’s Vascular Project Group. These priorities were converted into 4 questions:
In adults what are the effective risk assessment methods to identify:-
Q1. Patients with established clinical coronary vascular disease (CVD) [coronary artery disease, heart failure, peripheral vascular disease, cerebrovascular disease].
Q2. Patients who do not yet have established clinical CVD but are at very high risk of developing it over the next 10 years. These include patients with diabetes mellitus, familial dyslipidaemia and those considered to be at high global CVD risk (>20% over the next 10 years) on the basis of their risk factor profile.
Q3. Patients with an increased CVD risk in the population.
Q4. Patients considered to be hard to reach i.e. specific groups who may be less willing or able to engage with traditional healthcare delivery systems.
The evidence found in the literature searches consisted predominately of Level 2 (systematic reviews, cohort studies), Level 3 (observational studies) and Level 4 (expert opinion). Fifty seven papers were used to inform the report.
Q1 – General practice is considered the optimum setting for identifying patients with established CVD. Completeness of the data in disease registers is vital and the Quality and Outcomes Framework reviews and public health input are vital in ensuring this.
Q2 - A validation study of existing strategies for the diagnosis of diabetes indicated that further refinement is required before they can be used as the first step in screening for diabetes. There is increasing awareness that chronic kidney disease is associated with CVD. Evidence from cohort studies demonstrates the role of measuring glomerular filtration rate in vascular risk assessment.
Q3- There are many risk estimation tools in existence and there were several good quality evidence reviews on these tools. Risk prediction models are usually assessed using the criteria of calibration and discrimination. Discrimination can be illustrated by Receiver Operated Curves (ROC) which show discriminatory capacity over a range of possible thresholds. The most frequently used tool is the Framingham model that was recommended by the WAG Vascular Project Group and by other groups and guidelines. Framingham has however been the source of criticism and other tools such as ASSIGN, SCORE and QRISK have been developed; the most often cited problem with Framingham was its predictive accuracy and the lack of inclusion of criteria for deprivation. Tables 1 and 2 show comparisons and details of the risk tools, prepared from the literature. In order to address some of the problems with existing risk tools new risk factors have been trialled for inclusion, but high quality evidence for their effectiveness was not found in the literature searches (Table 3).
There are many guidelines that approach the issue of vascular risk assessment. A high quality systematic review appraised the published guidelines relevant to vascular risk assessment using the AGREE score, an accepted method for assessing guideline quality. Of the UK guidelines, those from the National Institute for Health and Clinical Excellence (NICE) and the Scottish Intercollegiate Guideline Network (SIGN) scored very high on the AGREE score, whilst the Joint British Societies, British Hypertension Society and Diabetes UK scored very low. Table 4 shows the factors to include for CVD risk from the 7 guidelines that had the highest AGREE scores. All guidelines recommend consideration of age, sex, smoking status, blood pressure, high density lipoprotein cholesterol (HDL-C) and total cholesterol.
The 2010 update to the NICE guidance on lipid management considered the use of the QRISK2 tool versus the Framingham model. The NICE Guidance Executive agreed that the Framingham risk equation should no longer be considered the equation of choice for the assessment of CVD risk but should be considered one of the possible equations to use. The Executive could not make a definite decision about the evidence for the superiority of QRISK over Framingham. In March 2010 Professor Roger Boyle, National Director for Heart Disease and Stroke recommended general practitioners (GPs) use QRISK to make CVD risk assessments. A comprehensive external validation study of QRISK and Framingham revealed that the QRISK score is more accurate in a UK population.
In 2009 the NHS Health Check Service was commenced in England. This vascular risk assessment programme invites all women and men between the ages of 40 and 74 years, free of diabetes and CVD and not receiving treatment for hypertension, who have not been previously identified for a health check. This policy of universal screening has large cost implications for any health service and there is literature that questions whether there are more cost effective approaches to risk assessment. One alternative method for risk prediction is pre-selection or targeted selection of patients using data routinely collected in GP practices. Economic modelling data, prepared for the NICE lipid modification guidance supports an approach that initially prioritises individuals using estimates of their CVD risk using data held in electronic medical records. An expert review that included modelling techniques indicated that the least efficient strategy for identification of treatment eligible patients was opportunistic assessment and the most efficient was prior estimation of CVD risk using data already held in electronic medical records. One of the review’s conclusions was that UK general practices have enough data to identify the patients who should be the highest priority for CVD risk factor assessment. Modelling and cohort studies on pre-stratification have also been reported using a database initially set up to record cancer patients in Norfolk. The results from seven screening strategies indicated that compared with universal screening a stepwise screening approach using a simple risk score incorporating routine data was an effective method to pre-select patients. A major limitation to the studies was that cost effectiveness analyses were not performed and there is also a need to adjust for standardised mortality rates.
Q4 – The searches revealed national guidance and systematic reviews that are aimed at determining interventions that are effective in identifying patients from hard to reach groups. These reviews discuss the problem with the Framingham equation of its inability to accurately predict risk in areas with high levels of deprivation. The ASSIGN and QRISK tools have been recommended for use in socially deprived populations, but high level evidence was not found to support their effectiveness. Methods of engaging hard to reach groups are being considered by another Public Health Wales group.
1 Introduction
Vascular disease (coronary heart disease, diabetes, stroke and renal disease) affects a large number of the population of the United Kingdom (UK) and in 2006 deaths from the disease in Wales were 11, 300.[1] People from deprived communities and ethnic minorities are selectively affected and vascular disease accounts for a large part of population health inequalities.
A risk factor may be defined as a characteristic of a person that is associated with an increased risk of developing a specific disease such as atherosclerotic cardiovascular disease (CVD). To be clinically relevant, it should be accepted as causal and modifiable, and a defined benefit should result from such modification (or if not modifiable it should enable identification of those at risk e.g. gender). Most risk estimation systems include age, gender, smoking, blood lipids, and blood pressure as their core variables. In this context, age is a measure of exposure time and not a risk factor as such.[2]
The UK National Screening Committee published guidance for vascular risk assessment (VRA) in March 2008 to inform and support a more structured approach for screening for vascular disease.[3]In April 2009 the NHS Health Check Service, a VRA programme formerly known as the vascular check programme, commenced in England.[4] All adults aged between 40 and 74 years, free of diabetes and CVD and not receiving treatment for hypertension, who have not been previously identified, will be invited to their surgery to attend a health check.[5] Implementing the Health Check Service programme has been estimated to prevent 9,500 myocardial infarctions and strokes per annum at a cost of £250 million per year.4 These figures are however mainly based on modelling studies with several assumptions and the real costs and benefits of vascular risk screening are unknown. [6],[7] It is therefore important to determine cost effectiveness of any vascular risk screening programmes.
The Welsh Assembly Government (WAG) is committed to improving cardiovascular health and healthcare in Wales. Standard 2 of the National Service Framework for Coronary Heart Disease states that everyone who is at high risk of developing CVD is offered multi factorial risk assessment.[8] It is accepted that a CVD risk in excess of 20% over 10 years constitutes a high risk. The Quality and Outcome Framework (QOF) of the GMS contract provides standards of care for patients with coronary heart disease (CHD), stroke, diabetes and hypertension. Furthermore it encourages screening for hypertension in people over the age of 45. There is no quality indicator for formal risk assessment.
A report from the Vascular Project Group under the aegis of WAG on vascular risk management for the people of Wales was published in 2010.[9] The authors of the report recognise that as well as lifestyle approaches to improve the risk of CVD in the population, it is necessary to identify those members of the population who are at risk of CVD. The document makes several recommendations for the basis of a strategy for risk prediction and a more comprehensive risk management approach so that CVD prevention is improved in Wales. The authors of the report suggest that the Framingham risk assessment model[10] has advantages over the QRISK tool.[11] The group recognises that the costs of implementing a national CVD assessment programme similar to the Health Check Service in England will be high and suggest that an alternative strategy including a pre-screening assessment be considered.
Other vascular risk prediction methods have been proposed, such as pre-stratifying people using data routinely collected in primary care, which may be more cost effective.[12] Public Health Wales has been involved in the work on vascular risk and was asked by WAG to review and assess the evidence on the various tools and methods for CVD risk prediction.
2 Aims
Rapid review of the evidence for the effectiveness of CVD risk prediction methods/tools.
3 Research questions
The research questions were formulated for the priorities listed in the WAG Vascular Project Group report.9
In adults what are the effective risk assessment methods to identify:-
Q1. Patients with established clinical CVD disease (coronary artery disease, heart failure, peripheral vascular disease, cerebrovascular disease).
Q2. Patients who do not yet have established clinical CVD but are at very high risk of developing it over the next 10 years. These include patients with diabetes mellitus, familial dyslipidaemia and those considered to be at high global CVD risk (>20% over the next 10 years) on the basis of their risk factor profile.
Q3. Patients with an increased CVD risk in the population.
Q4. Patients considered to be hard to reach i.e. specific groups who may be less willing or able to engage with traditional healthcare delivery systems.
The 4 questions were converted to the Patient, Intervention, Comparison and Outcome (PICO) format.[13] Throughout the document the term vascular risk applies to patients with coronary heart disease, stroke, transient ischaemic attack, type 2 diabetes mellitus peripheral arterial disease and chronic kidney disease.